oulu 2015 acta - search homejultika.oulu.fi/files/isbn9789526208374.pdf · viimeisessä osassa...
TRANSCRIPT
UNIVERSITY OF OULU P .O. Box 8000 F I -90014 UNIVERSITY OF OULU FINLAND
A C T A U N I V E R S I T A T I S O U L U E N S I S
Professor Esa Hohtola
University Lecturer Santeri Palviainen
Postdoctoral research fellow Sanna Taskila
Professor Olli Vuolteenaho
University Lecturer Veli-Matti Ulvinen
Director Sinikka Eskelinen
Professor Jari Juga
University Lecturer Anu Soikkeli
Professor Olli Vuolteenaho
Publications Editor Kirsti Nurkkala
ISBN 978-952-62-0836-7 (Paperback)ISBN 978-952-62-0837-4 (PDF)ISSN 0355-3213 (Print)ISSN 1796-2226 (Online)
U N I V E R S I TAT I S O U L U E N S I SACTAC
TECHNICA
U N I V E R S I TAT I S O U L U E N S I SACTAC
TECHNICA
OULU 2015
C 532
Mikko Leinonen
FINITE ELEMENT METHOD AND EQUIVALENT CIRCUIT BASED DESIGN OF PIEZOELECTRIC ACTUATORS AND ENERGY HARVESTER DYNAMICS
UNIVERSITY OF OULU GRADUATE SCHOOL;UNIVERSITY OF OULU,FACULTY OF INFORMATION TECHNOLOGY AND ELECTRICAL ENGINEERING,DEPARTMENT OF ELECTRICAL ENGINEERING
C 532
ACTA
Mikko Leinonen
C532etukansi.kesken.fm Page 1 Tuesday, May 12, 2015 12:55 PM
A C T A U N I V E R S I T A T I S O U L U E N S I SC Te c h n i c a 5 3 2
MIKKO LEINONEN
FINITE ELEMENT METHOD AND EQUIVALENT CIRCUIT BASED DESIGN OF PIEZOELECTRIC ACTUATORS AND ENERGY HARVESTER DYNAMICS
Academic dissertation to be presented with the assent ofthe Doctoral Training Committee of Technology andNatural Sciences of the University of Oulu for publicdefence in the OP auditorium (L10), Linnanmaa, on 26June 2015, at 12 noon
UNIVERSITY OF OULU, OULU 2015
Copyright © 2015Acta Univ. Oul. C 532, 2015
Supervised byDocent Jari JuutiProfessor Heli Jantunen
Reviewed byProfessor Stephen BeebyProfessor Pasi Kallio
ISBN 978-952-62-0836-7 (Paperback)ISBN 978-952-62-0837-4 (PDF)
ISSN 0355-3213 (Printed)ISSN 1796-2226 (Online)
Cover DesignRaimo Ahonen
JUVENES PRINTTAMPERE 2015
Leinonen, Mikko, Finite element method and equivalent circuit based design ofpiezoelectric actuators and energy harvester dynamics University of Oulu Graduate School; University of Oulu, Faculty of Information Technologyand Electrical Engineering, Department of Electrical EngineeringActa Univ. Oul. C 532, 2015University of Oulu, P.O. Box 8000, FI-90014 University of Oulu, Finland
Abstract
The main objective of this thesis was to use and combine Finite Element Method (FEM) and smallsignal equivalent circuit models in actuator and energy harvesting design and to study thedynamics of the said designs.
The work is divided into four different sections. In the first section, the small signal parametersare derived for a pre-stressed piezoelectric actuator using a series of measurements. In addition,the tunability of the resonance frequency using mass and series capacitors is investigated.
In the second section, a piezoelectric Fabry Perot Interferometer actuator is simulated usingFEM and the small signal parameters are derived using FEM simulations. The modelled resultsare compared with the actual measurements and the resonance frequency is found to differ by only0.8 percent from the measured values when the mirror is attached to the actuator.
In the third section a piezoelectric wide band energy harvester is developed with multiple beamtopology. Two different designs are compared, one produced using the conventional PZT-steelstructure and one with a PZT-LTCC structure.
The final section presents an FEM model for a shoe mounted energy harvester and concentrateson the modelling of walking dynamics in FEM. The simulation results are compared to actualmeasurements and the simulated power values are found to differ by only 7% when the cymbalstroke is below 1.3 mm. The generated model is also expandable to other types of energyharvesters and the methods developed can be used in a variety of different energy harvestingsimulations and harvester development.
The results show that the equivalent circuit approach together with FEM modelling is apowerful tool in the dynamics design of piezoelectric actuators and energy harvesters.
Keywords: actuator, energy harvester, FEM, piezoelectric
Leinonen, Mikko, Elementtimenetelmän ja vastinpiirien käyttö pietsosähköistenaktuaattorien ja energiankorjuukomponenttien dynamiikan suunnittelussaOulun yliopiston tutkijakoulu; Oulun yliopisto, Tieto- ja sähkötekniikan tiedekunta,Sähkötekniikan osastoActa Univ. Oul. C 532, 2015Oulun yliopisto, PL 8000, 90014 Oulun yliopisto
Tiivistelmä
Väitöstyön päätavoitteena oli yhdistää elementtimenetelmät (FEM) ja piensignaalimallit aktuaat-torien ja energiankorjuukomponenttien suunnittelussa ja tutkia niiden dynamiikkaa.
Työ on jaettu neljään eri osaan. Ensimmäisessä osassa piensignaalimallit johdettiin pietsosäh-köisestä aktuaattorista mittausten avulla. Lisäksi resonanssitaajuuden muuttamista tutkitaan mas-san ja sarjaan kytketyn kapasitanssin avulla.
Toisessa osassa simuloidaan pietsosähköistä Fabry Perot interferometria käyttäen elementti-menetelmää. Lisäksi komponentin piensignaalimalli luodaan käyttäen simulointimallia. Lopuksipiensignaalimallin ja prototyypin mittaustuloksia verrataan. Mallin resonanssitaajuus poikkeaamitatusta vain 0.8 %, kun aktuaattoriin on kiinnitetty peili.
Kolmannessa osassa kehitetään ja verrataan toisiinsa kahta erilaista laajakaistaista monipalk-kista pietsosähköistä energian korjuukomponenttia. Toinen komponenteista on toteutettu perin-teisellä PZT-teräs rakenteella ja toinen yhteissintratulla PZT-LTCC rakenteella.
Viimeisessä osassa luodaan simulaatio malli kenkään asennetulle cymbal tyyppiselle pietso-sähköiselle energian korjuukomponentille ja kävelyn dynamiikkaa tutkitaan. Luotua mallia ver-rataan prototyypin mittaustuloksiin ja simuloitu energian tuotto poikkeaa vain 7 % alle 1.3 mmpuristusliikkeellä.
Tulokset osoittivat, että piensignaalimallin ja elementtimenetelmän yhdistäminen on tehokasapu pietsosähköisten aktuaattorien ja energiankorjuukomponenttien dynamiikan suunnittelussa.
Asiasanat: aktuaattori, energiankorjuukomponentti, FEM, pietsosähköisyys
7
Acknowledgements
First of all, I wish to thank my supervisors Docent Jari Juuti and Professor Heli
Jantunen for supervising and supporting my work. Furthermore, I would like to
thank the colleagues at Microelectronics and Materials Physics laboratories for a
good working environment. Jaakko Palosaari, Maciej Sobocinski and Jari Juuti
have been of tremendous help in my work and for that I am grateful. Professor
A.E. Hill has also been very helpful in correcting my grammar in this thesis and
in our papers.
The Finnish Foundation for Technology Promotion, Wihuri Foundation,
KAUTE Foundation and GETA graduate school are acknowledged for financially
supporting this work.
Oulu, May 2015 Mikko Leinonen
9
List of abbreviations and symbols
α Overlapping factor
ε Permittivity
ξ Damping ratio ν Poisson’s ratio
ω Angular frequency AFM Atomic Force Microscope BW Bandwidth C Capacitance Cs Series capacitance CST Constant Strain Triangle CTE Coefficient of Thermal Expansion d Displacement
e error
E Young’s modulus
F Force FEM Finite Element Method FPI Fabry-Pérot Interferometer
fres Resonance frequency
GA Genetic Algorithm
IR InfraRed
k Spring constant keff Effective spring constant l Length of a beam LASER Light Amplification by Stimulated Emission of Radiation LCR Inductor, capacitor and resistor circuit LTCC Low Temperature Co-fired Ceramic LVDT Linear Variable Differential Transformer m Mass madded Added mass meff Effective mass MFC Macro Fiber Composite n Transformer turn ratio
NIR Near InfraRed
PZT Lead Zirconium Titanate Rin Input resistance
10
rm Mechanical resistive impedance representing mechanical losses
SOI Silicon On Insulator
SPICE Simulation Program with Integrated Circuit Emphasis
Vin Input voltage Vp-p Peak-to-peak voltage
x Displacement
11
List of original papers
This thesis is based on the following five original papers, which are cited in the
text by the given roman numerals.
I Leinonen M, Juuti J & Jantunen H (2005) Equivalent circuit based modification of a pre-stressed piezo actuator. Proc. 4th Int. Conf. on Smart Systems, Seinäjoki.
II Sobocinski M, Leinonen M, Juuti J & Jantunen H (2012) A Piezoelectric Active Mirror Suspension System Embedded Into Low-Temperature Cofired Ceramic. IEEE trans. on Ultrasonics, Ferroelectrics and frequency control 59(9): 1990-1995.
III Leinonen M, Palosaari J, Juuti J & Jantunen H (2011) Piezoelectric energy harvester for vibrating environments using multiple beam topology for wideband operation. AP XIX Proceedings 41, Helsinki.
IV Sobocinski M, Leinonen M, Juuti J & Jantunen H (2011) Monomorph piezoelectric wideband energy harvester integrated into LTCC. Journal Of European Ceramic Society 31(5): 789-794.
V Leinonen M, Palosaari J, Juuti J & Jantunen H (2014) Combined electrical and electromechanical simulations of a piezoelectric Cymbal harvester for energy harvesting from walking. Journal of Intelligent Material Systems and Structures, 25(4): 391-400.
In Paper I an equivalent circuit for a piezoelectric actuator is developed using
measurements and is then used to tune the resonance frequency of the actuator. In
Paper II a mirror suspension actuator is developed on LTCC for IR spectroscopy.
In this paper the FEM simulations are used as the main tool in the design. In
Paper III a wideband energy harvester is developed with 5 differently tuned
piezoelectric beams using a conventional PZT-Steel structure. In Paper IV an
improved wideband energy harvester is developed with LTCC technology in order
to improve the resonance frequency accuracy. In Paper V an energy harvester for
walking is modelled and simulated using FEM software and the results are
compared with actual measurements.
In Paper I the idea, measurements and the main part of the writing was done
by the author. In Paper II the idea, design and the modelling was done by the
author and the manufacturing and measurements were done by co-authors. In
paper III the idea, theory and design of the harvester as well as part of the analysis
of the results was
done by the author. The manufacturing and the measurements were done by the
co-authors. In Paper IV the idea of the energy harvester and the design as well as
the main part of the writing was done by the author. The manufacturing and the
measurements were done with the help of co-authors. In Paper V the development
12
of the simulation model, simulations and the main part of the writing were done
by the author and the idea and design of the harvester as well as the
measurements were done by the co-authors.
13
Contents
Abstract
Tiivistelmä
Acknowledgements 7
List of abbreviations and symbols 9
List of original papers 11
Contents 13
1 Introduction 15
1.1 Actuator and energy harvesting dynamics .............................................. 15
1.2 Scope and outline of the thesis ................................................................ 18
2 Small signal model parametrization for pre-stressed piezoelectric
actuator 19
2.1 The Butterworth-Van Dyke equivalent circuit ........................................ 19
2.2 Experimental development of the equivalent circuit parameters ............ 21
2.2.1 The design and manufacturing of the test device ......................... 21
2.2.2 Measurement results and the development of the
equivalent circuit model ............................................................... 22
3 FPI Actuator design 27
3.1 FPI mirror actuator .................................................................................. 27
3.2 The design of the FPI-actuator ................................................................ 27
3.3 The simulation model of the FPI-actuator ............................................... 28
3.4 The simulation results of the FPI-actuator .............................................. 30
3.5 The equivalent circuit parameters of the FPI-actuator ............................ 31
3.6 Verification of the equivalent circuit parameters of the FPI-
actuator using measurements .................................................................. 32
4 Energy harvester designs 35
4.1 Wide band energy harvester .................................................................... 35
4.1.1 PZT-steel wideband energy harvester .......................................... 39
4.1.2 LTCC wideband energy harvester ................................................ 40
4.1.3 Measurement results and comparison between the
harvesters ...................................................................................... 41
4.2 Modelling energy harvesters for walking ................................................ 47
4.2.1 The design of the shoe mounted energy harvester ........................ 48
4.2.2 The FEM model of the energy harvester ...................................... 49
4.2.3 The simulation results of the energy harvester ............................. 52
5 Conclusions 59
15
1 Introduction
Piezoelectric actuators, sensors and energy harvesters are widely used in different
applications throughout industry and are also encountered in many everyday tasks
in people’s lives. These piezoelectric components can be categorized as
transducers that either transform electrical energy into kinetic energy (actuators)
or vice versa (energy harvesters and sensors). The rate of this energy
transformation is defined by the mode of actuation, sensing or energy production.
These modes can be roughly divided into two operating groups as static and
dynamic operation modes. Static or quasistatic operation is encountered in
actuation and sensing but usually not in energy harvesting. Dynamic operation, on
the other hand, is used in all of the transducer applications. The dynamics of the
transducers play an important role in the design of the said components. For
example, the actuators and sensors in micropositioning systems should preferably
work outside resonance in order to maintain a linear frequency response and
avoid ringing. Energy harvesting applications, on the other hand, usually benefit
greatly from operating in resonance. Some actuator applications such as
ultrasonic actuators are also dependent on operation at resonance frequencies.
1.1 Actuator and energy harvesting dynamics
Piezoelectric actuators come in many sizes and can be used to drive bulky
equipment as well as micro- or even nano-scale objects. Applications where large
objects are actuated include, for example, telescopes such as in the Japanese
Subaru telescope where the mirror is actively suspended, thus avoiding the
disadvantages of a rigid mirror [1,2]. Similarly, a telescopic active hexapod
utilizes piezoelectric actuators [2,3]. In [4] a method was presented to replace the
conventional control surfaces with piezoelectric composite fibres (MFC)
embedded into an aircraft wing. A similar concept was presented in [5] with
adaptive control implemented using piezoelectric actuators. Piezoelectric
actuators are, however, more prominent at the smaller end of the size spectrum.
The atomic force microscope (AFM) and scanning probe microscope both utilize
piezoelectric actuators in the manipulation of the tip [2,6]. The resolution of these
devices is in the nanometre scale and with small scanning areas (5x5x1 µm) the
resolution can reach 0.1 nm in the XY-direction and 0.015 nm in the Z-direction
[7,8]. The nonlinear nature of the piezoelectric actuators means that, in
micropositioning, control systems are needed together with the sensors to
16
linearize the system. Closed loop control, however, requires that the dynamical
properties of the actuator are known in order to guarantee stability. Some open
loop control schemes are also viable; for example, current drive, where the effect
of the nonlinear capacitance of the piezoelectric component is avoided by using
charge as the input. This reduces hysteresis, which can be around 15 % without
any control [9], and requires no instrumentation other than metering of the current
[9,10]. For example, in [11], the initial hysteresis of 27 % was reduced to only
2 % by using current drive. In current drive the model for the piezoelectric
component is not needed, but can be included to improve performance [11].
However, in the case of both open (feed-forward) and closed loop control the
model needs to be known. The exception to this is the use of various adaptive
control systems, where the parameters of the model are not known a priori, but
rather generated iteratively during the operation of the device [12].
In some open loop control methods the system is modelled and then the
inverse of the model is applied as the controller, thus altering the dynamics and
nonlinearities of the system. This method relies on very accurate modelling of the
system and in some cases only the nonlinearities are modelled while preserving
the dynamical characteristics of the system [2,13]. The hysteresis can be modelled
with various methods such as the Preisach model and its variant Prandtl-Ishlinskii
models. These models have yielded good results in hysteresis cancellation: for
example an uncontrolled error of 17.5 % was reduced to 2 % using the Prandtl-
Ishlinskii model in [14].
Closed loop control, by its very nature, requires a feedback signal from the
output. For example, in micropositioning applications this signal derives from the
displacement of the actuator and in acoustic devices this can be the sound
pressure signal from a microphone or hydrophone. The feedback control requires
a high quality feedback signal because any interference and noise added to the
feedback signal is also treated as an input signal and is therefore usually passed
unattenuated to the output. In micropositioning applications the movements are
small, usually in the range of a few micrometres or even nanometres, and
measuring such small movements makes high demands on the instrumentation of
the system. Methods commonly used in conjunction with piezoelectric actuators
include Linear Variable Differential Transformer (LVDT), Light Amplification by
Stimulated Emission of Radiation (LASER), capacitive, strain gauge and
piezoresistive sensors [15-19]. In addition, the model of the system should be
known in order to control the system reliably, as is the case with most of the
control systems described.
17
The Finite Element Method (FEM) is traditionally used in mechanical
engineering to study the strain and stress in solid material. FEM is therefore well
suited for modelling piezoelectrical components and various examples of design
based on FEM modelling exist, for example in fuel injector design [20,21]. In
energy harvesting FEM modelling can be used to design structures where
analytical methods can be too difficult to realize. FEM modelling can also enable
optimization with methods such as genetic algorithms (GA) [22].
Small signal models have been traditionally used in electric circuit design,
but mechanical systems can be modelled using the same circuit components as in
electric circuit models. Piezoelectric components, having both electrical and
mechanical components, are therefore well suited for equivalent circuit analysis
where lumped circuit parameters represent the mass, spring and losses of the
component. The most classical equivalent circuit is the Butterworth Van Dyke
circuit [23]. This model is usable when the first resonance mode of the component
is under examination and the operation is considered linear. Nonlinearities such as
impact can be modelled, for example, by adding nonlinear components to the
circuit [24]. In this work, the small signal model is used to study the dynamics of
the actuators and to tune the resonance frequency. Nonlinear parameters are
encountered in the studied cymbal harvester, and can be used to explain the
nonlinear behaviour under high load conditions.
Harvesting energy from vibration is most efficiently done at the resonance
frequency of the piezoelectric component, and when designing a wideband energy
harvester the resonances have to be designed in such a way as to create a pass-
band. This task also requires a knowledge of the dynamics of the vibrating
piezoelectric components. This can be acquired by creating a small signal model
of the vibrating component. In particular, in multiple beam wideband energy
harvesters, the individual beams’ resonance frequencies have to be tuned in order
to create a wide pass-band. This task requires a knowledge of the beam’s
dynamics and this can be acquired by creating a small signal model of the beam.
With the aid of this model the dynamics can then be designed in such a way that
the required wideband operation is achieved. Multiple beam harvesters have been
manufactured using electromagnetic [25] and piezoelectric beams [26].
Piezoelectric bimorphs with varying piezoelectric layer thicknesses were used in
[27] to achieve the desired resonance frequency distribution. Beam length was
varied in [28] and masses were varied in [28,29].
18
1.2 Scope and outline of the thesis
In this work, small signal models are generated for a multitude of actuators and
energy harvesters. The emphasis is on the resonance frequency behaviour and
therefore the Butterworth-Van Dyke equivalent circuit is used as the model. The
parameters were obtained either by measurements or by FEM-simulations.
In Chapter 2, an equivalent circuit model is obtained for a pre-stressed
piezoelectric component. The pre-stressing makes it difficult to obtain the
parameters using analytical methods alone and thus they were obtained using a
series of measurements.
In Chapter 3, a Fabry-Pérot Interferometer (FPI) actuator is developed and
characterized. The dynamics of the mirror actuator are critical in interferometer
applications and therefore the dynamical model is required. The model was
obtained using FEM and then compared using measurements.
In Chapter 4, two different wideband energy harvesters are designed. The
principle of operation is dependent on the accurate control of resonance
frequencies in both the design and manufacturing phases. Actuators were
manufactured by two different methods and the measurement results were
compared to pinpoint design and manufacturing aspects which affect the
dynamical behaviour of the wideband energy harvester. In addition, a model for
an energy harvester which harvests energy from walking is developed. The
developed model is shown to accurately model the energy output from the
piezoelectric energy harvester when it is installed into the sole of a shoe. Also, the
dynamics of the energy harvester are examined, as the spring constant in this case
is highly nonlinear and the implications of this are further discussed.
The purpose of this work is to provide tools for small signal model creation
and show the importance of these models in actuator and energy harvester design.
Actuator dynamics can be tuned and evaluated before manufacturing and
previously manufactured components can be characterized and their small signal
model can be created using methods presented in this work. In energy harvesting,
better wideband energy harvesters can be manufactured by using small signal
models. The concept of spring constant can also be used in studying the nonlinear
behaviour encountered in high displacement shoe mounted piezoelectric energy
harvesters.
19
2 Small signal model parametrization for pre-stressed piezoelectric actuator
2.1 The Butterworth-Van Dyke equivalent circuit
Pre-stressed piezoelectric actuators have not been properly modelled or
parameterized before. Multilayer pre-stressed piezoelectric stack actuator was
modelled as a simple 2nd degree system and parameterized using step response in
[30], but bending pre-stressed actuators have not been properly parameterized.
The pre-stressing changes both the mechanical and electrical properties of the
actuator [31] and therefore parametrization should be done using measurements.
The most commonly used equivalent circuit is the Butterworth-Van Dyke circuit
[32] shown in Fig. 1. Other more complicated equivalent circuits have also been
developed [33-35], but the classical Butterworth-Van Dyke is the simplest and
easiest to implement. It consists of a voltage source Vin, the voltage source’s
output resistance Rin, the capacitance C of the piezo component, a transformer
with a turn ratio of n, and the mechanical losses which are modelled as rm. The
effective mass of the component is modelled as an inductor with an inductance of
m and the spring constant, k, is shown in this circuit as a capacitor. The electrical
domain in the left side of the circuit has only the capacitance of the piezo
component and the mechanical domain on the right side of the circuit is
comprised of an inductor, capacitor and resistor (LCR) circuit which effectively
determines the resonance frequency of the component. This is true when the
coupling between the electrical domain and the mechanical domain is weak i.e.
the transformer ratio is low, which is usually the case in piezoelectric devices.
20
Fig. 1. The classical Butterworth-Van Dyke equivalent circuit for piezoelectric
components [Paper I, published by permission of Frami OY].
The resonance frequency, when assuming lossless component i.e. rm=0, for the
circuit in Fig. 1. can be expressed as
(1)
and the transformer ratio by
(2)
where F is the force output of the component under Vin excitation. As can be seen,
the most obvious method to tune the resonance frequency is to alter the mass of
the component by adding mass, madded, to the point of the greatest displacement
under vibration. Then the effective mass becomes meff = m+madded .
When current drive is used, for example to reduce the hysteresis of the
component, the capacitance of the piezoelectric component couples to the
resonating LCR-circuit and the resulting resonance frequency is
(3)
As can be seen, the capacitance stiffens the component by adding to the spring
constant. Thus the new effective spring constant is
(4)
As the spring constant can be altered by the capacitance of the piezoelectric
component, the capacitance should be tunable. This, however, still requires the
21
current drive for the component and in some cases this is not viable. The same
capacitance tuning can be obtained with voltage drive by using the circuit shown
in Fig. 2. Now a capacitance Cs is connected in series with the actuator and in
effect prevents the voltage source from grounding the capacitance of the
piezoelectric component.
Fig. 2. The equivalent circuit for a piezoelectric actuator with a series capacitor
[Paper I, published by permission of Frami OY].
The equation for the resonance frequency is now
(5)
From this equation, it can be seen that the resonance frequency can be tuned
between the voltage driven case in equation (1) and current driven case in
equation (3). For example, if Cs is 0, the resonance frequency is given by (3) and
when Cs >>C, then resonance frequency is obtained by equation (1).
2.2 Experimental development of the equivalent circuit parameters
2.2.1 The design and manufacturing of the test device
The equivalent circuit was developed for a simple pre-stressed unimorph bender,
which was clamped from both ends as shown in Fig. 3. The dimensions of the
piezoelectric element were 5 x 34 x 0.25 mm. The material was PZT-5H from
Vin
1:n
C
m
1/k
CS
22
Morgan Electro Ceramics Inc. The piezoelectric disc, 35 mm in diameter, was
pre-stressed by printing 60 µm Ag-paste on one side and then firing at 850 °C.
The coefficient of thermal expansion (CTE) mismatch between the paste and the
PZT material created the pre-stress. The details of this process are presented in
[36] and [37]. The pre-stressed disc was then cut with a Nd:YVO4 laser and poled
in a 6 MV/m electric field at 100 °C for 30 min. A steel strip with dimensions of 5
x 38 x 0.05 mm was glued on the PZT beam and then fixed to clamps 0.7 mm
from the edges of the PZT beam. A mirror was attached to the centre of the beam
for subsequent displacement measurements using a Michelson interferometer
[38].
Fig. 3. The pre-stressed unimorph piezoelectric disc clamped from both sides [Paper I,
published by permission of Frami OY].
2.2.2 Measurement results and the development of the equivalent
circuit model
The measurements consisted of displacement and force measurements. The
displacement measurements were done with a Michelson interferometer detailed
in [38] and with a signal generator, Agilent 33120A. The force measurements
were performed using a Kyowa WGA-650B instrumentation amplifier and a
Kyowa LUB-20KB force sensor.
The transformer ratio, n, was obtained using force measurements. The
actuator was loaded with 1 N mechanical pre-stress by a force sensor and then
actuated with a varying voltage and the change in the force reading recorded. Due
to the low sensitivity of the force sensor, relatively high voltages had to be used.
Using equation (2) the n together with the error estimate was plotted in Fig. 4.
The error estimate was based on the ±0.01 N error of the force reading. The
transformer ratio, n, at 200 V was recorded as 2.8 mN/V [Paper I].
23
Fig. 4. The transformer ratio n and the error estimate versus the input voltage Vin
[Paper I, published by permission of Frami OY].
A displacement measurement was used to determine the spring constant, k. Under
static conditions k=nVin/x and by simply measuring the displacement x under Vin
excitation using the previously obtained n the k was calculated as 15.6 kN/m
[Paper I].
A resonance measurement was used together with equation (1) to determine
the effective mass, meff, which in this case is m, since madded was 0. The resonance
was obtained by using displacement measurement and sweeping the frequency.
The resulting resonance frequency was found to be 1053 Hz and, using equation
(1), the effective mass meff was obtained as 0.296 g [Paper I].
The capacitance of the piezoelectric component was measured to be 30 nF
and if the mechanical losses are omitted i.e. rm=0, then the equivalent circuit is
fully parameterized for studying the effects of added mass and an external
capacitor on the resonance frequency. The effect of mass was studied by adding
masses i.e. madded in 0.3 g increments, with starting point of 0.779 g i.e. the mass
of the bolt.. The resonance frequencies at each mass are shown in Fig. 5 where the
calculated values using the small signal parameters and equation (1) are also
plotted. As can be seen, the calculated values differed only slightly from the
0.002
0.0025
0.003
0.0035
0.004
0.0045
0.005
0 20 40 60 80 100 120 140 160 180 200
n
Voltage (V)
nError
24
measured values. The greatest relative error between the calculated and the
measured value was 2.5 % at 1.08 g of added mass, as seen in Table 1 [Paper I].
Fig. 5. The resonance frequency versus the added mass of the actuator [Paper I,
published by permission of Frami OY].
Table 1. The errors between the measured and the estimated resonance frequencies
[Paper I, published by permission of Frami OY].
added mass (g) 0.779 1.08 1.38 1.68 1.98
error (%) 0.6 2.5 1.4 0.3 -1.9
The effect of external capacitance with a simple voltage drive was studied by
measuring the resonance frequencies of the prototype and varying the external
capacitance, Cs, in the circuit shown in Fig. 2. Four different capacitors were used
in the measurements with capacitances 10, 22, 47 and 100 nF respectively and the
change in the resonance frequency is shown in Fig. 6 where the analytical results
are also shown using equation (5). The prototype had a madded of 0.779 g in the
measurements. The error of the frequency measurement was ±0.5 Hz, since the
minimum step size for signal generator was 1 Hz, and is shown in Fig. 6 as
vertical bars [Paper I]. The measured values closely followed the analytical values
400
450
500
550
600
650
2.01.91.71.51.31.10.90.7
Freq
uenc
y (H
z)
Mass (g)
Estimated resonance freq.Measured resonance freq.
25
and therefore, taking into account the added mass measurements, the small signal
equivalent circuit parameters were successfully obtained.
Fig. 6. The change in resonance frequency of the prototype versus the series
capacitance [Paper I, published by permission of Frami OY].
A pre-stressed piezoelectric component was manufactured and its small signal
model was characterized. When comparing measured and calculated resonance
frequencies between the component and its model, a good agreement was found.
The model, however, did not incorporate losses, which could therefore affect the
obtained simulated results. Nevertheless, as the results show, the absolute error
between the component and its model was 2.5 % at maximum and the omission of
rm seems to be a valid decision. The methods provided to parametrize the model
can be used in a variety of piezoelectric actuators. Furthermore, the frequency
tuning using the model is a useful tool for tuning vibrating piezoelectric
components. Series capacitance modulation, i.e. using a tunable capacitor in
series with the component, can be used to fine tune the resonance of the
component. Energy harvesting applications can benefit from this approach since
the frequency can be tuned during operation and thus maintain the optimal
harvesting frequency when using high Q-factor harvesters. A high Q-value in
energy harvesters is beneficial in terms of obtained energy, but also means that
26
the energy can only be harvested from a very narrow frequency band. Therefore
tunability of components in energy harvesting is an ongoing research endeavor.
Mechanical pre-stress devices have been proposed [39], also electrostatic springs
with a comb structure [40] and a piezoelectric method [41] where the vibrating
beam is bent by a secondary beam to change the shape of the vibrating beam.
Electrical tuning methods have been proposed, for example, by adding a shunt
capacitance [42] which is similar to the method proposed in this work. The
difference is that even with low load impedances the series capacitance can shift
the frequency and therefore provide frequency tuning by shielding the electrical
side of the small signal model from grounding. This was demonstrated using an
inherently low impedance voltage drive on an actuator and recording the change
of resonance frequency as shown in Fig. 6. The tunability of the component is
naturally dependent of n, as is shown in [43].
27
3 FPI Actuator design
3.1 FPI mirror actuator
In spectroscopy a Fabry-Perot-interferometer (FPI) may be used to filter desired
wavelengths from infrared spectra to determine, for example, the chemical
composition of a medicine. Tunability of this interferometer is desirable because
the filtered spectrum can then be swept and analyzed. An FPI-filter usually
consists of two partially reflecting mirrors with an air gap between the mirrors.
The air gap determines the filter frequency and in order to tune this frequency a
straightforward approach is to move one or both of the mirrors. Various actuating
methods have been used previously such as piezoelectric stack actuators [44-46],
thermal actuators [47] and electrostatic actuators [48]. The near infrared channel
(NIR) is between 700 – 1000 nm and therefore piezoelectric actuation is well
suited for the FPI-filter since the displacement range is ~1 µm which can easily
be achieved with piezoelectric materials. The state of the art FPI-filter, of which
the author is one of the inventors [46], is shown in [49, 50] where stack-type
piezo actuators are installed in the recesses on the side of the mirrors. The
drawbacks of the design are the cost of the actuators and the mechanical stresses
they introduce to the mirror, which therefore has to be thick in order to provide
stiffness to the structure. In this work, the stack actuators are replaced with piezo
beams, which results in reduced costs and allows the use of smaller and lighter
mirrors. The bending beams, however, can also induce some stresses to the mirror
and therefore flexural joints are introduced to enable the use of smaller and
thinner mirrors in the filter. The dynamical properties of the FPI-actuator are then
examined using both the FEM-method and the small signal method. The aim of
the work was to provide a more cost effective alternative to the state of the art
actuator with smaller and lighter mirrors.
3.2 The design of the FPI-actuator
The goal of this work was to develop an FPI-filter in a TO-3 package, which
meant that the size of the filter was constrained to a diameter of 15 mm with a 3
mm optical aperture. TO-3 package was set as a goal, since it is a standard
package used in optical measurement instruments. A three actuator structure was
used to provide mirror tilting properties for the filter. This enabled the zeroing of
28
the manufacturing-introduced optical gap variation, i.e. the mirror tilt. Fig. 7.
shows the layout of the filter. The actuators were implemented as beams, as
opposed to the stack actuators used previously [44-46]. This reduced the required
voltage and provided high displacements in a relatively compact size. The
material of the beams was PZ-29 from Ferroperm and the passive layer was
Heraeus Heralock 2000 zero shrinkage LTCC tape. This manufacturing process is
described in [51-52]. The beams were connected to the mirror holder using a
flexural joint. Flexural joints are frequently used in micropositioning stages [53-
55] and their design and theory can be obtained from [56-57]. Flexural joints were
used to minimize the stress induced in the mirror because stress can bend the
mirror and therefore degrade the optical performance of the filter. In cases where
the mirror bends, the optical gap between the mirrors varies and therefore the
effective aperture changes. The hinges also reduce the stiffness of the beams and
therefore increase the displacement of the mirror.
Fig. 7. The FPI-mirror actuator layout.
3.3 The simulation model of the FPI-actuator
The purpose of the modelling of the FPI-actuator was to minimize the stress
induced in the partially reflecting mirror attached to the mirror holder. Also, the
dynamics of the actuator were of interest because the settling time of a gap
adjustment determines the frequency sweep duration. In addition to the simulation
of the actuator dynamics, static steady state simulations were carried out in order
29
to determine the obtainable gap modulation i.e. the displacement of the mirror and
the stresses induced in the mirror and the flexural joints.
Comsol Multiphysics software (Comsol Inc., Burlington, MA) was used in
the FEM modeling. The model for the FPI-actuator was developed by modelling
only the beams and the mirror holder because the supporting rim structure was
assumed to be rigid. Therefore the beams were only modelled to have a fixed
constraint at the base, since the actuator was glued with cyanoacrylate onto a
platform. The active layer was modelled as PZ-29 with the material parameters
shown in Table 2, obtained from [58]. The passive layer was modelled as
isotropic material with a Young’s modulus E of 85 GPa, Poisson’s ratio ν of 0.24
and density of 2900 kg/m3 obtained from [59]. The material properties of the
mirror were obtained from [60] with E of 74 GPa and ν of 0.17. The meshed
geometry is shown in Fig. 8.
Table 2. The material properties of PZ-29, PZT-5H, steel and plastic cushions, [38,53]
[Paper V].
Material density SE11,22 SE
12,21 SE13,23,31,32 SE
33 SE44,55 SE
66 d31,32 d33 d15,24 ε11,12 ε33 E ν
kg/m3 1x10-12 m2/N 1x10-12 m2/N GPa
PZ-29 7460 17 -5.78 -8.79 22.9 54.1 45.6 -243 574 724 2440 2870
PZT-5H 7500 16.5 -4.78 -8.45 20.7 43.5 42.6 -274 593 741 3130 3400
Steel 7850 200 0.33
Plastic cushion 1150 2 0.4
Fig. 8. The meshed geometry of the FPI-actuator [Paper II, published by permission of
IEEE].
30
3.4 The simulation results of the FPI-actuator
The first simulation was a static displacement model where a potential of 100 V
was applied to all of the beams. The displacement was recorded from the edge of
the mirror holder. The resulting displacement was 1.87 µm. However, the actual
measured value was ~1 µm, using OFV-5000 vibrometer (Polytech GmbH,
Germany). This error could be explained partly by the poling of the actual
component but the bending of the structure could also be a factor since the
simulation model omitted the supporting rim structure completely. Some
deformation can occur at the base of the beams and was also not included in the
model. The same simulation was used to study the stresses induced in the hinges
under actuation. The von Mises stress levels were recorded as 37.2 MPa, as
shown in Fig. 9. This is still well below the fracture limit of 200 MPa for the
LTCC material used [58].
Fig. 9. The von Mises stress (MPa) in the hinge.
The dynamic simulations were done using the eigenfrequency analysis mode on
the Comsol Multiphysics software. The first resonance mode was located at 14.7
kHz, whereas the measured resonance was 13.02 kHz. Other resonance modes
that were simulated are shown in Fig. 10 [Paper II]. As can be seen, the first
resonance is the first mode of a simple beam and the second and the third
resonance modes are the tilting resonance modes of the whole structure. The
fourth resonance mode is the second bending mode of a beam and it does not
include the tilting of the mirror. Analysing these results, it can be deduced that the
operating frequencies for this actuator are well below the first resonance of 14.7
31
kHz. However, some tilting of the mirror was observed at the 21.2 kHz resonance
and this has to be taken into account when driving the actuator in continuous
operation because mirror tilt could degrade the FPI-operation.
Fig. 10. The first resonance, second, third and fourth resonances at a) 14.7 kHz
b) 21.2 kHz c) 71.9 kHz and d) 77.4 kHz respectively [Paper II, published by permission
of IEEE].
3.5 The equivalent circuit parameters of the FPI-actuator
The dynamical properties of the FPI-actuator were studied with a mechanical
equivalent circuit similar to Fig. 1. In this case, however, only the resonance
behaviour was studied and therefore the equivalent circuit was reduced to
incorporate only the inertial mass, m, and the spring constant, k. The resonance
frequency of this system is stated in equation (1). The total effective mass of the
system was m=meff+mm, where meff is the effective mass of the moving parts of
FPI-module and mm is the mass of the attached mirror. The parameters for
equivalent circuit were obtained by simulating the structure with varying mirror
weights. For example if we have two different mirrors with weights m1 and m2,
the effective mass can be obtained as
32
(6)
where f1 is the resonance frequency with mass m1 and f2 is the resonance
frequency with mass m2. The spring constant can be then obtained as
2 (7)
The structure was simulated using a parametric sweep for the mirror density, in
order to change the mirror mass. The simulation was carried out as an
eigenfrequency simulation to obtain the resonance frequencies. Then using
Equation (6), the mean effective mass meff was obtained as 35.6 mg, using mirror
masses ranging from 14 mg to 61 mg. [Paper II]. Now, with the effective mass,
meff , using equation (7), the spring constant, k, was obtained as 326 kN/m [Paper
II]. With these values it is now possible to assess the dynamical performance of
the FPI-actuator using different sizes and with mirrors of different materials.
3.6 Verification of the equivalent circuit parameters of the FPI-
actuator using measurements
The resonance frequency of the prototype was measured both with and without a
mirror attached. The attached mirror had a diameter of 4 mm and a thickness of
0.7 mm. The mass of the mirror was 15 mg and thus the density was calculated as
1704 kg/m3. The measured resonance frequency without the mirror was 13.02
kHz and with the mirror it was 11.31 kHz. Thus the shift in resonance frequency
was 13.1 %. The calculated resonance frequencies using meff, k, and Equation (7),
were 14.7 kHz and 12.5 kHz without and with the mirror, respectively. The
change in resonance frequency was therefore 15.5 %. As can be seen, the absolute
values between the calculated and measured values differed by 12.9 % between
the mirror-less cases and by 10.5 % when the mirror was attached. The measured
change in resonance frequency after attaching the mirror was 1.71 kHz and the
simulated change was 2.2 kHz and 2.4 % in percentage points. However, when
the model is applied to the measured case, i.e. using the meff of 35.6 mg obtained
from simulations and calculating the new k using Equation (7), the new k is
obtained as 339 kN/m. With this value we can calculate the new resonance
frequency with the mirror as 11.4 kHz. The error is only 90 Hz or 0.8 % when
compared to the measured value of 11.31 kHz. The reason for the difference
between the simulated and measured results can be attributed to slight variations
33
in geometry and to the fact that the material parameters of both the passive layer
and the active layer change when PZT material is sintered together with LTCC
[31].
The results demonstrate that by using the equivalent circuit in conjunction
with FEM-modelling it is possible to characterize the linearized model, and to
use the model as a quick tool for estimating the contribution of the mirror’s
weight to the resonance frequency of the component. The procedure to do this is
first to use FEM-simulations to obtain an average meff and then calculate the k
using the measured resonance frequency of the FPI-actuator without the mirror,
and finally, using Equation (7), obtain the resonance frequency with the mirror
attached.
An FPI-actuator was developed using beam type actuators instead of stack
actuators, which are commonly used in FPI filters. The beam structure was
proven to provide the required displacement of 1.87 µm in simulation and a
measured value of 1 µm. This was obtained by using flexural hinges to reduce the
spring constant and also to minimize the stresses induced in the mirror. The
structure differs from state of the art structures by providing a more cost efficient
and lighter alternative while having a similar performance. A simple spring mass
model was parameterized using FEM-simulations to find the effective mass of the
actuator. The model then may be used as a quick tool for estimating frequency
shifts when the mass of the mirror is changed. The model also enables the
estimation of settling time using standard second order system equations, which is
also an important tool in actuator design.
35
4 Energy harvester designs
In many applications a power source is needed to provide energy for the sensors,
data processing and communication. The usual power sources are batteries,
accumulators and wired power lines. However, batteries need exchanging,
accumulators need recharging and power lines are costly to build and not so
convenient in many cases. To tackle these problems various energy harvesting
methods have been developed. Energy harvesting can be defined as collecting
small amounts of energy from the environment surrounding the component in
question. Energy sources typically include solar [61-62], vibration [63], pressure
[64-66], magnetic and RF-fields [67-69], chemical [70-71] and temperature [72-
74]. Piezoelectric materials are well suited for energy harvesting from vibration
and motion in general. Furthermore, the size of piezoelectric harvesters is usually
very compact and they are relatively easy to manufacture compared to
electrostatic and electromagnetic components.
4.1 Wide band energy harvester
In industrial environments vibration is usually encountered in machinery with
rotating parts. This provides a natural source of energy for sensors. When dealing
with rotary machinery, however, the majority of the vibration is present at the
angular frequencies at which the machines are rotating. This means that the
energy harvesters have to be tuned to the same frequency. Piezoelectrical
components have a high mechanical Q-value and therefore tuning the resonance
frequency can be difficult. For example, to match the rotational frequency of a
machine, the narrow bandwidth of the harvester requires a precise match to the
machine’s frequency. A wideband solution where the Q-value is reduced permits
frequency tuning which is much less critical. Also, if the machine has some
variation in the rotation speed, the wideband energy harvester is able to continue
harvesting the energy provided the frequency of the machine stays within the
bandwidth of the harvester. By the same argument, if the source of the vibration is
random, a wideband energy harvester is able to harvest more energy compared to
a conventional harvester.
Multiple beam wideband piezoelectric energy harvesters include [26,75-76]
and also an electromagnetic harvester in [25]. MEMS implementations include a
4 beam SOI structure [77] and a 2 beam micromachined structure [78]. Other
methods have been utilized such as coupled resonance structures [79]. Most
36
recent wideband harvesters use bistable operation, either with magnets [80],
buckled beams [81] or springs [82]. These energy harvesters are nonlinear as
opposed to multiple beam resonators which are usually linear in behaviour. The
multiple beam topology was chosen, because it can provide higher bandwidths by
increasing the number of beams. The drawback of this design is its sensitivity to
manufacturing tolerances, as will be shown. A method to incorporate
manufacturing tolerances into design parameters will be presented in the work.
The beam lengths are calculated using a simple recursion instead of the analytical
methods presented in [83-84]. The harvesters presented here were manufactured
using laser processing as in [26]. A PZT-Steel harvester consisted of a PZT layer
glued to a steel layer and a novel PZT-LTCC harvester consisted of a PZT layer
and an LTCC layer co-sintered together and then laser cut into shape in order to
minimize dimensional errors. This novel technique has not been previously
proposed in energy harvesters. The two harvesters are compared with the effect of
manufacturing tolerances in mind. The design of the wide band energy harvester
Two different wideband energy harvesters were designed and manufactured. A
multiple beam topology was chosen to accomplish the wideband operation. In this
beam topology the energy harvester comprises of multiple piezoelectric beams,
each with a different resonance frequency
In multiple beam topology, the individual resonance frequency of the n:th
beam is given as
, (8)
As the length of the beam is increased, the spring constant, kn , decreases and the
mass, mn , increases. The absolute values for the spring constant and mass are not
needed if only the changes in resonance frequency are required rather than the
actual resonance frequencies. Simple relations for the masses and spring constants
can be used to accomplish this resonance frequency tailoring. It is known from
beam theory that the spring constant of a beam has the relation
∝ (9)
, where l is the length of the beam. The mass of the beam has the relation
∝ (10)
Substituting these into equation (8), the relation for the resonance frequency is
∝ (11)
37
Now the difference in resonance frequencies between two beams using the beams’
damping ratio, ξ, and overlapping factor, α, is generated in this work as
, , , (12)
If the overlapping factor is 2, then the relative difference between the beams is
exactly the same as the relative bandwidth of the beam, namely 2ξ , which
coincides with the half power point in the frequency response. If an energy
harvester with beams separated with overlapping factor 2 is made, a flat power
spectrum is generated assuming that the power generated by each beam is
superimposed. When the overlapping factor is 0, the resonance frequencies are all
equal and the total bandwidth of the harvester is the same as that of a single beam,
i.e. 2ξ. The power, however, is the total sum of individual beams’ power. The
ideal value for the overlapping factor is therefore 2 with respect to wide band
operation and when using n beams the total bandwidth is 2ξn. A more general
expression for the bandwidth BW, with the inclusion of the overlapping factor, α,
is
2 1 (13)
Now, in order to design the wideband harvester, a recursion based on equations
(11) and (12) is developed as
, (14)
With this equation it is possible to design the resonance frequencies of the beams
to satisfy the bandwidth requirements. The starting frequency has to be either
calculated or simulated using FEM software to obtain the desired centre
frequency. It is also possible to alter the resonance frequency by adding mass to
the beams, but in order to taper the resonance frequencies, different masses are
required for individual beams.
During manufacturing, different errors are introduced into the dimensions of
the beams. The thickness, width and length of the beams can be altered by these
errors. For example, the error in beam length can be analyzed as follows. Using
equation (12) we get
,
,1 (15)
And by using equation (11), we get
1 (16)
38
now if absolute error e is introduced into ln we get
,
,~ 1 (17)
,where e/ln is the relative error. This equation shows the effect of error on the
resonance frequency shift. To compensate for this we define new overlapping
factor αc which can be obtained by setting
1 1 1 (18)
and solving for αc
1 1 1 (19)
where α is the desired overlapping factor. For example, we may set α as 2 and ξ as
0.025 and assume a positive relative error of 0.01 i.e. the actual length of the
beam is 1 % longer than the desired length and therefore the resonance frequency
is lower than expected. The compensated overlapping factor is obtained as 2.84.
This means that the beams are shortened to compensate for the error and the
resulting frequency shift is the same as with an overlapping factor of 2 without
any error present. In real world cases the error’s sign is rarely known a priori and
therefore a worst case scenario can be analyzed. If the sign of the error is opposite
between beams, then 2e can be used as a value for the error. In this case, if the
overlapping factor is chosen as 2, the worst case scenario widens the gap between
the two resonances in such a way that the ripple exceeds the 3dB limit. For
example, if we have a 1 % error in the length of the beams, we can use -2 % as
the relative error (2 % shorter beam than anticipated) with an overlapping factor 2
and ξ again as 0.025. Using equation (19) we get the compensated overlapping
factor as 0.32, which differs greatly from the original overlapping factor of 2.
This demonstrates the high sensitivity of the frequency tuning of multibeam
energy harvesters.
In this case, only the error in the length of the beam was analyzed, but by
using beam equations, the errors for width and thickness can also be analysed
using similar methods to those presented here.
39
4.1.1 PZT-steel wideband energy harvester
A five beam energy harvester was designed and manufactured. A circular closed
shape was chosen in order to facilitate encapsulation of the component. The
outline of the harvester is shown in Fig. 11.
Fig. 11. a) The dimensions and b) the manufactured prototype [Paper III, published by
permission of Suomen Automaatioseura].
The active material for the prototype was PZT-5H from Morgan Electroceramics.
The material parameters can be seen in [85]. The passive layer was steel and the
thicknesses of the active and passive layers were 375 µm and 150 µm
respectively. The length of the first beam was chosen as 22 mm, in order to fit as
long beams as possible inside the 35 mm piezo disc with sufficient support area.
The width of the beam was also maximised to 4 mm in order to improve the
power output of individual beams. It is important to maximize the beam length in
order to reduce the resonance frequency of the harvester because the natural
frequency is high for beams of this size. When the base frequency of the harvester
is made as low as possible, it is easier to reduce it further to the desired level by
adding mass at the tip of the beams. The design parameters for calculating the
length of the beams were 0.025 for the damping ratio, ξ, and 0.5 for the
overlapping factor, α. Damping ratio was chosen using a priori experience of
vibrating piezoelectric beams. A low value for the overlapping factor was chosen
in order to ensure proper overlapping and therefore higher power output at the
expense of the bandwidth, since hand gluing and placement generate higher
tolerances. These tolerances could be estimated as 100 µm for placement, but
glue thickness was not estimated, but could have been measured using a series of
40
cross-sectional cuts. The resulting beam lengths were 22, 21.9, 21.7, 21.6 and
21.4 mm.
4.1.2 LTCC wideband energy harvester
In addition to the traditional PZT-steel unimorph wideband energy harvester a
PZT-LTCC monomorph structure was also manufactured in order to test whether
the more precise monolithic process improved the frequency performance of a
wideband harvester. The active material of the harvester was PZ29 (Ferroperm
A/S, Denmark) instead of the PZT-5H used for the PZT-steel harvester. The
passive material was Heraeus Heralock 2000 LTCC tape. The main difference
between the previous prototype and the LTCC embedded prototype was in the
manufacturing of the layered structures. In the previous prototype each layer was
cut with a laser and then glued together. The gluing, which was done by hand,
added errors in the positioning of the layers in the XY-plane and also in the Z-
direction as the glue layer thickness was not precisely controlled. In the LTCC
prototype, however, the positioning was done using a jig and the layers were
sintered together so that a glue layer with arbitrary thickness was not present. The
prototype is shown in Fig. 12.
Fig. 12. The PZT-LTCC harvester without (middle) and with lid (right) [Paper IV,
published by permission of Elsevier Ltd].
The harvester was designed with three beams instead of five as in the PZT-steel
prototype. The beam lengths were calculated using equation (14) and using the
starting length of 18 mm, the overlapping factor, α, of 1 and the damping factor, ξ,
41
of 0.025. Higher overlapping factor for this harvester was chosen as much higher
dimensional accuracy was expected for the monolithic structure. This resulted in a
length series 18, 17.78 and 17.56 mm. Beam width was 3 mm and the layer
thicknesses were 375 and 270 µm for the PZT and LTCC layers, respectively. The
difference between the beam lengths was 220 µm and for the PZT-steel prototype
it was 150 µm. The larger overlapping factor eased the manufacturing of the
harvester since the required length difference was greater than with a smaller
value of α. This is important in structures where the beam lengths are restricted as
in this case where the 25 mm disc diameter restricts the beam length compared to
the previous PZT-steel prototype where a 35 mm disc was used. Smaller disc
diameter was used because of material availability and better manufacturability of
the smaller disc.
4.1.3 Measurement results and comparison between the harvesters
The wide band energy harvesters were measured both as an actuator and a
harvester. The actuator measurements were done using a vibrometer OFV-5000
(Polytech GmbH, Germany) for the displacement measurements. A piezo stack
Piezomechanik Pst 150/7/160 VS12 was used to generate the desired acceleration
for the energy harvesting measurements. The energy harvesting electronics are
shown in Fig. 13 where a common rectifier is used together with a capacitor (1
µF) as the energy storage unite unit.
Fig. 13. The schematic of the energy harvesting electronics [Paper III published by
permission of Suomen Automaatioseura].
42
Actuator measurements were performed in order to assess the resonance
frequencies of the beams. Fig. 14 shows the frequency response curve of the PZT-
steel harvester under 10 Vp-p excitation. It can be observed that beam number 5
was not working, thus making the energy harvester effectively a 4 beam harvester.
Furthermore, the order of the resonances was 1-2-4-3 as opposed to the designed
order of 1-2-3-4. Fortunately this does not weaken the performance since the
spacing of the resonances is still acceptable. Beams number 4 and 5 were located
side by side and since the amplitude of beam number 4 was also attenuated as in
the case of beam 5, it can be said that same problem could be affecting for both
beams. This can be for example the glue layer between the steel and PZT, which
by altering the total thickness locally changes the resonance frequency of the
beams located in the said area. The resonance frequencies for the beams 1, 2, 3
and 4 were 695, 700, 724 and 712 Hz, respectively [Paper III]. According to
equation (12), resonances 695, 703.7, 712.5 and 721.4 Hz were obtained. When
comparing this series with the measured values, errors for the measured and
calculated frequencies were obtained as 0.5 %, 1.6 % and -1.3 %. The first
calculated frequency was assumed to be 695 Hz and the error was therefore 0 %.
As can be seen, the relative errors in the frequencies were small but could still
result in the beams switching places in the frequency domain. In this prototype
the resonance frequencies for beams 3 and 4 were very close to the designed
values as they have switched places, but this was not according to design. The
displacement amplitudes of the resonances were similar for beams 1, 2 and 3, but
beam 4 was attenuated to roughly half of the other beams’ displacement.
43
Fig. 14. The Frequency response of the PZT-steel energy harvester under 10 Vp-p
sinusoidal actuation [Paper III published by permission of Suomen Automaatioseura].
Similar measurements were performed for the PZT-LTCC prototype and the
resulting frequency response curve is shown in Fig. 15. As can be seen, the order
of the resonances was as desired 1-2-3 and the respective resonance frequencies
were 1109, 1137 and 1155 Hz [Paper IV]. The calculated values using equation
(12) were 1100, 1137 and 1165 Hz. Therefore the errors were 0.82, 0 and 0.86 %
for the frequencies. Thus, the errors were essentially half of those of the PZT-
steel prototype. Further insight into the effect of the overlapping factor is gained
when the frequency differences are studied. The differences fres,n-fres,n-1 for the
PZT-steel model were 5, 24, -12 Hz. The calculated values, however, were 8.7,
8.8 and 8.9 Hz. Now the errors for the differences were 74 %, 172 % and 234 %.
These values differ greatly from the absolute errors which were at maximum
1.6 %. The same measured differences for the PZT-LTCC prototype were 28 and
18 Hz and the calculated values were 37 and 28 Hz. Now the errors are 24 and
36 %. As the required frequency difference is larger due to the overlapping factor
α, the relative error also becomes lower. Therefore larger overlapping factors
should be used even though the power output is somewhat reduced because this
ensures that wideband operation is maintained even though errors are introduced
in the manufacturing process. This can however result in bandpass ripple
exceeding the 3 dB level and if that is not acceptable, then lower overlapping
factor should be used. This is a design choice that should be considered when
44
facing high manufacturing tolerances. When comparing the magnitudes of the
resonances between Figs. 14 and 15, it can be noted that for the PZT-LTCC
prototype the magnitudes are in descending order, which is characteristic of beam
resonances in actuation.
Fig. 15. The PZT-LTCC harvester under 1 Vp-p sinusoidal actuation [Paper IV,
published by permission of Elsevier Ltd].
The wideband operation of the harvesters was measured using a frequency sweep
and connecting the harvester into the circuit shown in Fig. 13. In this
measurement, the harvester was attached to a shaker and the frequency of the
shaker was varied while the amplitude remained constant. For the PZT-steel
prototype a 6 µm peak-to-peak displacement was used for the actuation. With
constant displacement, the acceleration amplitude ranged from 36 to 76 m/s2as the
frequency was swept from 550 to 800 Hz. The voltage of the capacitor was
recorded and the energy content of the capacitor was then calculated from the
voltage. The resulting frequency response is shown in Fig. 16. The centre
frequency was 717 Hz, with a 3 dB bandwidth of 54 Hz [Paper III]. The
improvement of the bandwidth when comparing it to a single beam, with the
designed Q-value of 20, is 50 %. Using measured Q-value of 24.6, the
45
improvement was 85 %. For the PZT-LTCC prototype the frequency response is
shown in Fig. 17. In this measurement, the amplitude was 1 µm, as the frequency
was higher than in the PZT-steel case. This resulted in accelerations between 20
and 33 m/s2. The centre frequency was 1147 Hz and the bandwidth was 62 Hz
[Paper IV]. Therefore the actual improvement in the bandwidth compared to a
single beam was 5.4 times larger i.e. 540 % since the single beam’s measured
bandwidth was 11 Hz. The measured Q-value of the single beam was 103 on
average and the value used in the design phase was 20. In comparison, the PZT-
steel prototype had an average Q-value of 24.6 which deviated only 23 % from
the value used in the calculations. The high Q-value of the PZT-LTCC prototype
is also seen in the Fig. 17 as the pass-band of the harvester had a considerable
ripple which should not exist with such a low overlapping factor. For the PZT-
steel prototype the pass-band notch was not due to a high Q-value, because the Q-
value was close to that of the value used in the calculations, but rather due to the
low output of beam number 4, evident in Fig. 14.
Fig. 16. The buildup voltage and energy of the capacitor on PZT-steel
energy harvester [Paper III published by permission of Suomen Automaatioseura].
46
Fig. 17. The buildup voltage of the capacitor of the PZT-LTCC harvester, the voltage
across the optimal load resistance and the energy of the capacitor.
As the PZT-LTCC prototype showed, the Q-value can differ greatly from design
to design. This affects the resonance frequency placement and if the error in the
Q-value estimation is large, the pass-band of the energy harvester will generate
ripple. Another important design decision is the choice of manufacturing process.
As was shown, the less precise manufacturing process where layers were glued by
hand in the PZT-steel prototype led to large errors in the resonance frequencies of
the beams. In contrast, the LTCC process proved to be very good in
manufacturing precise geometries and therefore the resonance frequencies of the
beams were very close to the designed values. The third important aspect is the
choice of the overlapping factor as it can to some extent alleviate the tolerances in
manufacturing, especially when using a large number of beams. This is of course
at the expense of bandwidth and energy, since a higher overlapping factor
increases the bandwidth but decreases the power and by using a lower
overlapping factor the effect is the opposite. The error analysis showed the high
sensitivity of length variations to the resonance frequency, and this can be used as
a design tool in the selection of manufacturing process and overlapping factor.
The performance of the harvesters could be improved by introducing mass at the
tip of the beams, because that lowers the resonance frequencies and also improves
the power output of the energy harvester [86]. In this work, masses were not
47
introduced, because the focus was on the frequency tuning by length variation of
beams and other variables were kept to a minimum.
The small signal analysis proved to be a useful tool when designing wideband
multibeam harvesters, since together with simple beam equations, process
tolerances and material Q-value information it could be used in designing the
beam resonance frequencies which fulfilled the specified frequency range
properties. In [26] a similar laser processed structure was developed and the
claimed improvement of the bandwidth was 200 % compared to that of a single
beam for a 6 beam structure and 33 % per beam. The results were 50 % and
540 % for the PZT-steel harvester and PZT-LTCC harvester, respectively.
Improvements per beam were 12.5 % and 180 %. A MEMS implementation in
[77] claimed resonance frequencies between 237 – 244.5 Hz with four beams,
which roughly translates to a relative bandwidth of 3.12 %, whereas the PZT-steel
prototype and the PZT-LTCC prototype obtained 7.5 % and 5.4 %, respectively.
Besides performance enhancements, the PZT-LTCC prototype provides an
integrated package for the harvester and also enables the integration of
harvesting, and sensor electronics into the harvester itself. This then enables very
robust harvesters for environmentally demanding applications, since LTCC is
both a circuit board and a package at the same time.
4.2 Modelling energy harvesters for walking
Wearable electronics have been developed to be used, for example, in firefighters’
outfits and in military attires. Various sensors embedded in shoes, gloves and
other wearable items need power, and batteries or accumulators are usually used.
However, the problem with these is the weight and the need to recharge or replace
them. Energy harvesting is a natural solution for wearable electronics. A wide
range of solutions exist for generating power from the movement of the body
such as harvesting the heel strike of the shoe [87], the pressure of the foot [88],
the movement of the knee joint [89], the straps of the backpack [90] and the heat
from the body[91]. In this work, however, the pressure and heel strike of the foot
were used as the energy source. Energy harvesting from walking presents many
challenges. The low operation frequency of ~ 1 Hz limits the use of resonating
structures and therefore the coupling of energy from the mechanical domain into
the electrical domain in piezoelectric materials. Also, the harvester should not
hinder the movement or provide discomfort to the user of the device. The energy
density of the harvester device should also be maximized since the space
48
available is limited. Furthermore, when embedding energy harvesters into shoes,
the force experienced by the harvester is not a simple step function, but rather has
a defined shape, which affects the power output and also the FEM modelling of
the device.
In this chapter, the performance of a cymbal harvester is studied using FEM-
modelling. The mechanical input into the harvester is measured and the results are
used as an input to the FEM simulator. The mechanical input in this case is the
force experienced by the harvester when mounted inside the sole of a shoe during
walking. FEM modelling provides a tool for understanding the behaviour of the
piezo electric harvester under high stresses during walking. The harvester in
question is a cymbal type transducer, which is typically used as an actuator [92-
93]. FEM modelling is a commonly used tool in cymbal [93-98], and also in
harvester design. Sinusoidal excitation was used in [99-100] with a relatively high
excitation frequency of 100 Hz. The power output of simple piezoelectric discs
was studied in [101]. FEM and analytical methods was used in [102], to study the
effect of an additional steel layer attached to the piezoelectric layer in a cymbal
configuration. In [103], circumferential slots were incorporated into the endcaps
of the cymbal and FEM together with analytical methods was used as a tool.
However, the accurate simulation of cymbals subjected to the actual walk profile
has not previously been studied. The incorporation of arbitrary waveforms,
nonlinear mechanics and large displacements into a single cymbal energy
harvester simulation is a novel approach which gives valuable information about
cymbal energy harvesters under high load conditions.
4.2.1 The design of the shoe mounted energy harvester
The shoe mounted piezoelectric harvester was designed to be embedded into the
sole of the shoe. Furthermore, the purpose of the study was to create a FEM-
model of the cymbal structure and to model the energy generation from walking.
Since the force experienced by the harvester is not sinusoidal, a transient model
with a relevant force profile is needed. Also, energy harvesting simulations
require the combination of both the mechanical and electrical domain simulations.
This model addresses all these requirements. The basis of the cymbal structure
was a PZT-5H disc 35 mm in diameter into which brass connector rings were
glued together with convex steel endcaps. The endcaps were pressed into shape
using a mould. This process was done manually and therefore the endcap shape
was not entirely uniform. The cymbal and its dimensions are shown in Fig. 18.
49
Fig. 18. The cymbal structure and the individual parts of the shoe mounted
piezoelectric harvester [Paper V, published by permission of SAGE publishing].
4.2.2 The FEM model of the energy harvester
The profile of the endcap was measured using a micrometer screw gauge and one
half of the profile was used in the model by mirroring it into the axial geometry.
This introduced some errors into the model since the endcap geometry was not
modelled in all quarters of the actual geometry. The measured profile is shown in
Fig. 19.
Fig. 19. The measured profile of the endcap used in model and measurements
[Paper V, published by permission of SAGE publishing].
50
The geometry of the model was designed to be parametric where all of the
geometric properties were added as variables that could be modified during
simulation passes. The shape of the endcap was modelled as a normalized shape
vector ranging from 0 to 1. The concavity and radius together with other
geometric parameters defined the actual points of the endcap. This design
provided a powerful tool for optimizing cymbal designs as in [22]. The meshed
geometry of the axial-symmetric design is shown in Fig. 20 together with plastic
cushions at the top and bottom of the cymbal. The plastic cushions were
incorporated into the model because the cymbal was installed into a plastic holder
which was then installed into a shoe sole. The force of the heel pressure was
therefore conveyed into the cymbal through these plastic cushions in a controlled
manner. The measurement system is presented in detail in Paper V.
The material parameters for the piezoelectric material, steel and the plastic
cushions are shown in Table 2. These values were obtained from the
manufacturer’s catalog [85] for the PZT. The steel’s and nylon plastic cushion’s
values were obtained directly from the Comsol software (Comsol Multiphysics
4.2 a). As can be seen from Fig. 20, triangular elements were used. The order of
the elements was quadratic and therefore the accuracy of the simulations was
adequate compared, for example, to constant strain triangle (CST) elements which
cannot properly calculate the stresses inside the triangle. Furthermore, the
geometric nonlinearities setting was applied for the simulations. This setting is
also known as the large deformation setting in some other FEM softwares. This is
necessary because the deformation levels are large in the simulations; the applied
stroke for the cymbal in the simulations was 1.5 mm, which is near the point
where the endcap touches the piezo disc.
Fig. 20. The meshed axial-symmetric geometry of the cymbal [Paper V, published by
permission of SAGE publishing].
51
The lower plastic pad was fixed from the bottom as a fixed boundary condition
and the top pad was in turn actuated with the input force. The force experienced
by the harvester was measured from the prototype during actual walking. This
force waveform was then used as an input for the top pad as a force type
boundary condition in the transient simulations. The force measurement data was
filtered to remove high frequency content in order to increase the step size in
simulations and therefore shorten the total simulation time. This filtered force
profile is shown in Fig. 21. As the results show, the pace of walking was roughly
1 Hz. This is also shown by [104-106] to be a good average pace for walking.
Fig. 21. The force input for the shoe mounted energy harvester at 0.5 mm stroke
[Paper V, published by permission of SAGE publishing].
The electrical boundary conditions were set as ground for the bottom electrode of
the PZT disc and the top electrode was set as a terminal connection for an
external SPICE circuit which consisted of a simple load resistor. This external
circuit was simulated concurrently with the FEM simulations using the same
Comsol software. The SPICE simulations provided an easy method for
calculating the power dissipated by the external load resistor. Furthermore, the
optimal load resistance was easily obtained using parametric simulations where
the resistance was swept.
52
4.2.3 The simulation results of the energy harvester
The first simulation was carried out as a static displacement vs. force simulation
where the stroke was applied to the top pad of the energy harvester and the
resulting force on the same surface was recorded, as shown in Fig. 22. The same
curve was also generated by measuring the prototype and these results are also
shown in the figure. As can be seen, the simulated values largely agreed with the
measured values although some differences appeared. Namely, in the area
between 0.6 to 1.6 mm the simulated forces were below the measured values and
also the nonlinearity was greater in the simulated curve than in the measured one.
The maximum error occurred at 1.1 mm with 8.8 % error. It has to be noted also
that the most significant contributors to this displacement vs. force curve were the
endcaps of the component, because the material of the piezoelectric disc deforms
very little compared to that of the endcaps. Therefore the mechanical properties of
the endcaps will have the most effect on the total performance of the harvester, as
will be shown later. The simulation type can also explain the difference between
measured and simulated results, since the model of the endcap was only one half
of an endcap mirrored to generate an axial symmetric geometry. From Fig. 22, the
average spring constants for the measured and simulated cases were obtained as
34.1 kN/m and 30.4 kN/m, respectively. As the nonlinearity of the curve is quite
strong a more general nonlinear temporal spring constant can be obtained by first
fitting a curve for the force vs. displacement curve. The force in the case of the
simulated model is
2.014 ∙ 10 4.114 ∙ 10 4.0808 ∙ 10 1.497 (20)
, where d is the displacement of the endcaps and the F is the reactive force on the
pad. The temporal spring constant is by definition the derivative of F with respect
to d as
6.042 ∙ 10 8.228 ∙ 10 4.0808 ∙ 10 (21)
This function reaches its minimum of 12.8 kN/m at 0.68 mm and its maximum of
63.7 kN/m at 1.6 mm when the displacement is restricted to 0.4 to 1.6 mm as in
the measurements. As Fig. 23 shows, the spring constant is highly nonlinear,
especially in the modelled case.
53
Fig. 22. The displacement vs. force of the cymbal actuator [Paper V, published by
permission of SAGE publishing].
For the measured harvester the force as a function of displacement was
1.137 ∙ 10 2.122 ∙ 10 3.955 ∙ 10 1.194 (22)
and the spring constant
3.411 ∙ 10 4.244 ∙ 10 3.955 ∙ 10 (23)
The minimum spring constant was located at 0.62 mm with a value of 26.4 kN/m.
The maximum at 1.6 mm resulted in 59 kN/m. When comparing the simulated
and measured spring constants, the measured spring constant had a higher
minimum and slightly lower maximum, therefore the simulated spring constant
exhibited a higher nonlinearity compared to the measured case. The minimum
occurred roughly at the same displacement. Furthermore, when comparing the
maximum deviation from the average spring constant obtained from the linear
fitting, the simulated case deviated 33.3 kN/m and the measured case deviated
24.9 kN/m. This shows that the average spring constant cannot be used in cases
where large deformations are encountered such as in a cymbal under high loads.
The spring constants in equations (21) and (23) are plotted in Fig. 23.
54
Fig. 23. The spring constant versus displacement of the measured and simulated
end cap of the shoe mounted piezoelectric harvester.
The force applied to the top of the end cap is transformed by the end cap into
radial stress in the piezoelectric disc. This transformation was simulated with a
similar simulation as in Fig. 22, where the force versus displacement was
modelled. In this case, however, the force was used as an input and the
displacement of the end cap and the von Mises stress inside the piezoelectric disc
was simulated. These results are shown in Fig. 24. As can be seen, at 30 N the
stress induced into the piezo disc started to drop. When comparing figures 23 and
24, it can be seen that the spring constant started to rise after 0.8 mm and at the
same displacement the change in stress growth in Fig. 24 started to decrease.
Therefore the drop in stress relaying properties of the end cap can indeed be
attributed to the nonlinear spring constant. Deformations of the end cap under
different loads are shown in Fig. 25. As can be seen, the deformations are
noticeable and at 30 N a recess had formed in the middle of the end cap. This
deformation therefore stiffened the structure giving rise to the highly nonlinear
spring constant shown in Fig. 23.
55
Fig. 24. Displacement and piezo stress as a function of force of the cymbal harvester
[Paper V, published by permission of SAGE publishing].
Fig. 25. The end cap shape under different loads [Paper V, published by permission of
SAGE publishing].
The power generation simulations were done using the force profile in Fig. 21.
The simulation was carried out as a transient simulation with the force applied to
56
the top pad. A similar sine signal was also used for comparison, with a peak-to-
peak value adjusted to match the walk profile. The results are shown in Fig. 26.
Fig. 26. The power generation with the walk and sine force profile under 17 N peak
load with optimal 2.63 MΩ load resistance [Paper V, published by permission of SAGE
publishing].
The load resistance was adjusted to 2.63 MΩ which was measured to be the
optimal resistance [Paper V]. The optimal resistance was also simulated using
transient simulation with the walk profile. In that case the optimal resistance was
3.0 MΩ which is very close to the measured value and therefore the use of the
measured value is also justified in the simulations. These results are shown in Fig.
27. As can be seen in Fig. 26, the walk profile produced a large peak when the
heel compressed the cymbal. When the heel was in the static pressure mode as
seen in Figs. 21 and 26 during 0.2 to 0.4 seconds, the power decays to zero and
when the heel rises and the cymbal experiences decompression the power rises
again to produce a similar peak but only smaller. When comparing the walk
profile with the sine profile, it can be seen that the compression cycle produced
roughly a 4.5 times higher peak compared to the sine profile [Paper V]. The
decompression cycle, however, produced a similar peak to the sine peak. The
compression cycle’s rise time was 60 ms and in the decompression cycle it was
184 ms. The shorter rise time explains the higher power output during the
compression cycle and also why the sine signal produced such a low power
57
output. The difference in power levels is also seen in Fig. 27, where the walk
profile produced ~100 µW more power in the 2.63 MΩ load resistor, which is
167 % higher power generation compared to sine signal. This result emphasizes
the importance of correct input signal modelling in energy harvesting simulations
where the input signal deviates from a sine signal.
Fig. 27. The power vs load resistance of the walk and sine input [Paper V, published by
permission of SAGE publishing].
The average power generations were measured using 12 different stroke values
ranging from 0.5 mm to 1.6 mm. The same displacements were also used in
simulations. In these simulations the profile in Fig. 21 was scaled to match the
force profile at the given stroke. The load resistance was adjusted to 2.63 MΩ in
both the measurements and simulations. Both results are shown in Fig. 28, where
the average power versus stroke distance is shown. The simulations were carried
out as transient simulations as in the case of the simulations depicted in Fig. 27
and the average power was calculated from one walk cycle and the result was
plotted. As can be seen from Fig. 28, the simulated values agree well with the
measured values up to 1.2 mm displacements. At 1 mm displacement, the
measurements show a deviation, which is probably caused by a measurement
error. The power output for the simulated model peaked at 1.3 mm with 640 µW.
The measured power peaked at 1.5 mm with 780 µW. The error between the
58
simulated and measured power below 1.3 mm was 7.1 % with the exception of
1 mm displacement where the error was 14.7 %. Above 1.3 mm the simulated and
measured values diverged and the simulated values saturated. As discussed
earlier, the saturation derives from the nonlinear spring constant of the end caps.
The same saturation did not occur until 1.5 mm displacement in the case of the
measured values. Fig. 23 shows that the measured spring constant exhibited
considerably greater linearity compared to the simulated case. This can also
explain the different saturation behaviour of the measured cymbal compared to
the simulated model. Therefore, when generating a model for a cymbal energy
harvester, a comparison between the spring constants is encouraged in order to
gain insight into establishing reliable models for energy harvesting simulations
with high deformations.
As the results showed, the developed model accurately modelled the
behaviour of the cymbal piezoelectric harvester. The accurate input signal and
geometry modelling, together with the combination of mechanical and electrical
simulations, generated a useful tool for cymbal type energy harvester design. This
model can be extrapolated into other topologies and operational environments and
thus provides added value to large stroke energy harvester designs.
Fig. 28. The average measured and simulated power versus displacement of the
cymbal harvester [Paper V, published by permission of SAGE publishing].
59
5 Conclusions
The main scope of this thesis was to develop methods to combine the equivalent
circuit analysis and finite element methods to improve the design process of
actuator and energy harvester dynamics. In addition, a new FPI structure was
developed to provide more cost effective alternative to the state of the art FPI
actuator. FEM analysis also provided further insight of the cymbal energy
harvester under large displacements during walking. The small signal parameters
were obtained either from measurements or from the FEM model. The small
signal model of the component could then be used in estimating resonance
frequencies of beams either in FPI actuators or wideband energy harvesters. In
addition, the nonlinear spring constant of a cymbal energy harvester was
discussed and its effect on the saturation behaviour was investigated in relation to
stroke displacement or input force.
The Butterworth-Van-Dyke equivalent circuit was developed for a pre-
stressed piezoelectric actuator. The small signal parameters were obtained from a
series of measurements. The model was tested against the component with
resonance measurements with varied added masses. The model deviated only
2.5 % at maximum. As the equivalent circuit reveals, the series capacitance can be
used in tuning the resonance frequency of the actuator. This was tested both with
the actual component and the small signal model and the calculated results
deviated only 14 % from the measured values. It was also noted, that by adding
a series capacitor into the piezoelectric component, the internal capacitance is
decoupled from the input/output impedance of the external electronics. This
provides a method to tune the resonance frequency of the energy harvester by
electrical means in low load impedance situations.
An FPI actuator was designed and simulated with PZ-29 as the active
material and LTCC as the passive material. The design featured a flexural hinged
mirror holder which alleviated stress in the mirror. This stress can bend the mirror
and therefore degrade FPI performance by affecting the local gap length. The
stress levels were also studied in the hinge area made of LTCC material and they
were confirmed to be well below the fracture limit. The reduction of stress levels
induced into the mirror enables the use of thinner and lighter mirrors which then
improves the transient properties of the FPI-actuator. In this work, the dynamical
parameters, namely the spring constant, k, and the effective mass, meff, of the
actuator were obtained using simulations. Eigenfrequency simulations were
carried out with a series of added masses. Resulting resonance frequencies were
60
then used in calculating the effective mass, meff , and the spring constant. The
small signal model was then compared to measurements where a mass was added
to the mirror holder. The resonance frequencies between the model and the actual
component differed only 0.8 % when mirror was attached, thus validating the
small signal approach. The results also showed that a small TO-3 sized FPI-
actuator can be made with beam structure instead of the state-of-the-art stack
actuators.
Two different types of wideband energy harvesters were designed to collect
energy from a vibrating environment and one cymbal type energy harvester to
collect energy from walking. A conventional PZT-Steel energy harvester was
developed and characterized. This harvester contained five differently tuned
beams, with the purpose of generating a series of resonances that are tuned in a
controlled manner to provide the wideband operation. The resonance frequencies
were tuned by controlling the beam length and thus the effective mass and the
spring constant of the beam. This approach proved effective as the difference
between the measured and calculated resonance frequencies was 1.6 % at
maximum, although two beams had switched places when put into order of their
resonances.
The second wideband harvester was a PZT-LTCC energy harvester which had
only three beams. This manufacturing process was more controlled because the
layers were stacked using a jig instead of manual alignment as in the PZT-Steel
case. The beams were tuned similarly to the PZT-Steel case apart from the
overlapping factor, α, which was 1 instead of 0.5 for the PZT-Steel case. The
results showed that the resonance frequencies differed only 0.86 % from the
calculated values which is nearly two times better than the PZT-Steel case. The
resulting frequency response of the energy harvester, however, exhibited ripple
which was an indicator of an incorrect damping factor, i.e. Q-value, assumption.
Surprisingly, the actual Q-value of the beam was five times that of the designed
value. This manifested itself as a pass-band ripple for the energy harvester.
Nevertheless, the PZT-Steel prototype did not suffer from this, since the designed
and measured Q-values were close. This proved that the overlapping factor and
Q-value have to be carefully decided and measured in the design process. In
addition, the manufacturing process has a very large influence on the resonance
frequencies, because the tolerances of the process can affect the geometry of the
component and therefore the resonance frequency. This was shown by an error
analysis, where the error on the beam length was studied. A method was
developed to incorporate this error into the overlapping parameter selection.
61
Small signal modelling when used in conjunction with correct Q-values, can
be used to tailor the bandwidth of multibeam harvesters. The design parameters
have to be selected according to manufacturing tolerances to optimize the
bandwidth. This was demonstrated using two different manufacturing processes
and the results gave valuable information about the tailoring of the design
parameters.
The energy harvester for walking was developed using a cymbal harvester.
The shape of the endcaps was measured and these results were then input into the
FEM model. As the harvester was designed to be embedded into the sole of a
shoe, the actual force profile of walking was measured and again inputted into the
model. This enabled FEM simulations with a high accuracy in simulating the
power levels obtained from walking. The almost completely parameterized model
provides a powerful tool for further optimizing cymbal type energy harvesters.
The model was first verified by carrying out a force vs. displacement simulation
as well as a similar measurement with the actual prototype. The results differed
by only 8.8 %, which was caused largely by the difference in the endcaps’ spring
constant. The spring constants were compared between the simulated and the
measured cases and the simulation model exhibited larger nonlinearity compared
to the measured component. This proved that the endcap modelling differed from
the actual component. However, as the power generation simulations and
measurements were carried out, the simulated model followed closely the values
of the measured component up to 1.3 mm displacement. After that the model
started to saturate. The measured component exhibited similar saturation, but at
larger displacements. Again, the stiffening of the structure was found to be the
cause of this behaviour. The endcaps’ ability to transfer input force into radial
stress in the piezoelectric disc showed a maximum at around 30 N after which the
stress started to drop and therefore the output energy of the piezoelectric
component saturated. This was a result of the stiffening of the endcaps under high
loads. The presented simulation tool provides a powerful method for optimizing
the endcap geometry to combat the stiffening encountered in this work.
This work presented methods to combine finite element techniques and small
signal analysis to improve the design process of actuator and energy harvester
dynamics. The small signal parameter, k, was also used in explaining the
saturation behaviour of the cymbal energy harvester, where it was found to be
nonlinear. This work presented novel methods to tailor the bandwidth of
wideband energy harvesters according to manufacturing tolerances and material
properties, namely Q-factor. The simulation model of a cymbal type energy
62
harvester used in a high stroke energy harvesting environment is a novel
contribution to energy harvesting. This work also proved that the accurate
modelling of the input signal together with time based transient modelling is an
essential part of high stroke simulations. The generated model is also expandable
to other types of energy harvesters and the methods developed can be used in a
variety of different energy harvesting simulations and harvester development.
63
References
1. Subaru telescope (2013). URI: http://www.subarutelescope.org. Cited 2013/12/12. 2. Juuti J, Leinonen M & Jantunen H (2008) Piezoelectric and Acoustic Materials for
Transducer Applications : chapter Micropositioning editori:Ahmad Safari & E. Koray Akdogan: 319–340.
3. PI (2001) Micropositioning, nanopositioning, nanoautomation solutions for cutting-edge technologies. PI catalog.
4. Li M, Yuan J, Guan D & Chen W (2011) Application of piezoelectric fiber composite actuator to aircraft wing for aerodynamic performance improvement. Science China Technological Sciences. 54(2): 395–402.
5. Suleman A & Costa AP (2004) Adaptive control of an aeroelastic flight vehicle using piezoelectric actuators. Computers & Structures 82(17-19): 1303–14.
6. Giessibl FJ (2003) Advances in atomic force microscopy. Rev Mod Phys 75(3): 949–982.
7. Veeco Intruments Inc (2003) SPM Training Notebook 004-130-00 Revision E. 8. Veeco Instruments Inc (2004) Dimension 3100 Manual Revision D. 9. Ang WT, Garmón A, Khosla PK & Riviere CN (2003) Modeling rate-dependent
hysteresis in piezoelectric actuators. In: Proc 2003 IEEE/RSJ International Conference on Intelligent Robots and Systems, IEEE, Piscataway, USA: 1975–1980.
10. Comstock RH (1981) Charge control of piezoelectric actuators to reduce hysteresis effects. U.S. Patent 4,263,527.
11. Ronkanen, P. Kallio, P & Koivo, H.N (2002) Current Control of Piezoelectric Actuators with Power Loss Compensation. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Lausanne, Switzerland, October pp. 1948–1953.
12. Åstöm KJ & Wittenmark B (2008) Adaptive Control:Second Edition, Dover Books. 13. Ru CH, Sun L & Kong MX (2005) Adaptive inverse control for piezoelectric actuator
based on hysteresis model. In: Proc 4th International Conference on Machine Learning and Cybernetics, IEEE, Piscataway, USA: 3189–3193.
14. Sun L, Ru C & Rong W (2004) Hysteresis compensation for piezoelectric actuator based on adaptive inverse control. In: Proc 5th World Congress on Intelligent Control and Automation, IEEE, Piscataway, USA: 5036–5039.
15. Dinulovic D & Gatzen HH (2006) Microfabricated inductive micropositioning sensor for measurement of a linear movement. IEEE Sens J 6 (6): 1482–1487.
16. Furuta A, Munekata M & Higuchi T (2002) Precise positioning stage driven by multilayer piezo actuator using strain gauge. Jpn J Appl Phys 41: 6283–6286.
17. Harb S, Smith ST & Chetwynd DG (1992) Subnanometer behavior of a capacitive feedback, piezoelectric displacement actuator. Rev Sci Instrum 63 (2): 1680–1689.
18. Sacconi A, Picotto GB & Pasin W (1999) The IMGC calibration setup for microdisplacement actuators. IEEE Trans Instrum Meas 48(2): 483–487.
19. Leinonen M, Juuti J & Jantunen H (2007) Interface circuit for resistive sensors utilizing digital potentiometers. Sensors and Actuators A:Physical. 138(1): 97–104.
64
20. Ferrari, A & Mittica, A (2012) FEM modeling of the piezoelectric driving system in the design of direct-acting diesel injectors. Applied Energy 99: 471–83.
21. Mehlfeldt, D, Weckenmann, H & Stohr G (2008) Modeling of piezoelectrically actuated fuel injectors. Mechatronics 18: 264–72.
22. Poikselkä K, Leinonen M, Palosaari J, Vallivaara I, Haverinen J, Röning J & Juuti J (2015) End cap profile optimisation of piezoelectric Cymbal actuator for quasistatic operation by genetic algorithm, Journal of Intelligent Material Systems and Structures, published online 2015/1/14, DOI: 10.1177/1045389X14566526.
23. Van Dyke, K.S (1928) The Piezo-Electric Resonator and Its Equivalent Network. Proceedings of the Institute of Radio Engineers 16(6): 742–764.
24. Wurpts, W & Twiefel, J (2012) Analysis of ultrasonic vibro-impact systems with equivalent circuits and the harmonic balance method. Sensors and Actuators A 200: 114–22.
25. Sari I, Balkan T & Kulah H (2008) An electromagnetic micro power generator for wideband environmental vibrations. Sensors and Actuators A 145-146: 405–13.
26. Rguiti, M Hajjaji, A D’Astorg, S Courtois, C & Leriche, A (2013) Elaboration and characterization of a low frequency and wideband piezoceramic generator for energy harvesting. Optical Materials 36(1): 8–12.
27. Xue, H Hu, Y & Wang, Q-M (2008) Broadband piezoelectric energy harvesting devices using multiple bimorphs with different operating frequencies. IEEE Trans. Ultrason.,Ferroelectr. Freq. Control 55: 2104–8.
28. Shahruz, S M (2006) Design of mechanical band-pass filters for energy scavenging. J. Sound Vib. 292: 987–98.
29. Ferrari, M Ferrari, V Guizzetti, M Marioli, D & Taroni, A (2008) Piezoelectric multifrequency energy converter for power harvesting in autonomous microsystems. SensorsActuators A 142: 329–35.
30. Hunstig, M & Hemsel T (2010) Parameter Identification and Model Validation for the Piezoelectric Actuator in an Inertia Motor. Journal of the Korean Physical Society 57(4): 952–4.
40. Palosaari J, Juuti J, Heinonen E, Moilanen P & Jantunen H (2009) Electromechanical Performance of Structurally Graded Monolithic Piezoelectric Actuator. Journal of Electroceramics 22(1): 156–162.
41. Berlincourt DA, Curran DR & Jaffe H (1964) Piezoelectric and Piezomagnetic Materials and Their Function in Transducers. in W.P. Mason ed., Physical Acoustics, Academic, NY, 1(A): 169–270.
42. Ballato A & Ballato J (1995) Modern Piezoceramic Equivalent Networks. Ultrasonics Symposium, Proc. IEEE 1: 683–686.
43. Adriaens JMTA, Koning WL & Banning R (1997) Design and Modelling of a Piezo-actuated positioning mechanism. Decision and Control, Proc. of the 36th IEEE Conference on. 2: 1978–1983.
44. Goldfarb M & Celanovic N (1997) Modeling piezoelectric stack actuators for control of micromanipulation. IEEE Contr. Syst. Mag. 17: 69–79.
65
45. Juuti J, Jantunen H, Moilanen VP & Leppävuori S (2005) Poling conditions of pre-stressed piezoelectric actuators and their displacement. Journal of Electroceramics 15: 57–64.
46. Juuti J, Jantunen H, Moilanen VP & Leppävuori S (2006) Manufacturing of Pre-stressed Piezoelectric Actuators by Post-fired Biasing Layer. IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency control 53(5): 838–46.
47. Moilanen H & Leppävuori S (2001) Laser Interferometric Measurements of Displacement-field Characteristics of Piezoelectric Actuators and Actuator Materials. Sensors and Actuators A 92: 236–334.
48. Eichhorn, C Goldschmidtboeing, F & Woias, P (2008) A frequency tunable piezoelectric energy converter based on a cantilever beam. Proc. PowerMEMS 2008+ microEMS2008 (9–12 November 2008, Sendai, Japan) 309–12.
49. Lee, K B Lin L & Cho, Y H (2008) A closed-form approach for frequency tunable comb resonators with curved finger contour. Sensors Actuators A 141: 523–9
50. Peters, C Maurath, D Schock, W & Manoli, Y (2008) Novel electrically tunable mechanical resonator for energy harvesting. Proc. PowerMEMS 2008+ microEMS2008, (Sendai, Japan, 9–12 November 2008) 253–6.
51. Muriuki, M. G. (2004) An Investigation into the Design and Control of Tunable Piezoelectric Resonators. Ph.D. Dissertation, University of Pittsburgh, PA.
52. Shu, Y. C Lien, I. C Wu, W. J (2007) An Improved analysis of the SSHI interface in piezoelectric energy harvesting. Smart Materials and Structures 16: 2253–64
53. Reay N, Ring J & Seaddan R (1974) A tunable Fabry–perot filter for the visible. J. Phys. E Sci. Instrum. 7(8): 673–677.
54. Atherton P & Reay N (1981) A narrow gap, servo-controlled tunable Fabry–perot filter for astronomy. Mon. Not. R. Astron. Soc. 197: 507–511.
55. Ollila J, Juuti J, Leinonen M & Saari H (2008) FI20085947 Fabry-Perot interferometer and method for its manufacture. Finnish patent.
56. Yu B, Pickrell G, Woong KJ & Wang A (2004) Thermally tunable extrinsic Fabry–perot filter for white-light interferometry. Frontiers in Optics.
57. Patterson J (1997) Micro-mechanical voltage tunable Fabry–perot filters formed in (111) silicon. Nasa Langley Research Center, Hampton, WV, nasa technical paper 3702.
58. Saari, H Aallos, V-V Akujärvi, A & Antila, T (2009) Novel miniaturized hyperspectral sensor for UAV and space applications. Proc. SPIE7474, Sensors, Systems, and Next-Generation Satellites XIII.
59. Mäkeläinen, A Saari, H Hippi, I Sarkeala J & Soukkamäki, J (2013) 2D Hyperspectral Frame Imager Camera Data in Photogrammetric Mosaicking. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences XL-1/W2: 263–7.
60. Heinonen E, Juuti J & Jantunen H (2007) Characteristics of piezoelectric cantilevers embedded in LTCC. J. Eur. Ceram. Soc. 27(13–15): 4135–4138.
66
61. Sobocinski M, Zwierz R, Juuti J, Jantunen H & Golonka L (2010) Electrical and electromechanical characteristics of LTCC embedded piezoelectric bulk actuators. Adv. Appl. Ceramics 109(3): 135–138.
62. Zhang Z & Hu H (2009) Flexural Mechanism Design Analysis for a New Piezoelectric Inchworm Actuator. Measuring technology and mechatronics automation, ICMTMA ‘09 International conference on 1: 98–101.
63. Furukawa E, Mizuno M & Doi T (1995) Development of a flexure-hinged translation mechanism driven by two piezoelectric stacks. JSME International Journal, Series C: Dynamics, Control, Robotics, Design and Manufacturing 34(4): 743–748.
64. Yang R, Jouaneh M & Schweizer R (1996) Design and characterisation of a low-profile micropositioning stage. Precision Engineering 18: 20–29.
65. Paros JM & Weisbord L (1965) How to design flexure hinges. Machine Design 37: 151–156.
66. Lobontiu N (2003) Compliant Mechanisms: Design of Flexure Hinges. CRC Press. 67. Data matrix from ferroperm website URI: http://www.ferroperm-piezo.com/ Cited
2013/12/10. 68. Roellig M, Schubert L, Lieske U, Boehme B, Frankenstein B & Meyendorf N (2010)
FEM assisted development of a SHM-piezo-package for damage evaluation in airplane components. in 11th. Int. Conf. Thermal. Mechanical and Multiphysics Simulation and Experiments in Micro-Electronics and Micro-Systems (EuroSimE): 1–9.
69. Shin-Etsu Chemical Co. Material datasheet URI: http://www.shinetsu.co.jp/en/products/pdf/garasu.pdf Cited 2013/10/12.
70. Green MA (2004) Third generation photovoltaics: advanced solar energy conversion. Springer Verlag.
71. Randall JF (2005) Designing indoor solar products. John Wiley & Sons. 72. Beeby SP, Tudor MJ & White NM (2006) Energy harvesting vibration sources for
microsystems applications. Measurement Science & Technology 17: 175–195. 73. Cunefare, K.A Skow, E.A Erturk, A Savor, J Verma, N & Cacan M.R (2013) Energy
harvesting from hydraulic pressure flunctuations. Smart Materials and Structures 22(2): 25–36.
74. Pfenninger, A Vogel, R Koch, V.M & Jonsson M (2014) Performance Analysis of a Miniature Turbine Generator for Intracorporeal Energy Harvesting. Artificial Organs 38(5): 68–81.
75. Deterre, M Lefeuvre, E & Dufour-Gergam, E (2012) An active piezoelectric energy extraction method for pressure energy harvesting. Smart Materials and Structures 21(8)
76. Mi, M Mickle, M.H Capelli, C & Swift, H (2005) RF Energy Harvesting with Multiple Antennas in the Same Space. IEEE Antennas and Propagation Magazine 47(5):100–6.
77. Moghe, R Divan, D & Lambert, F (2011) Powering Low-Cost Utility Sensors using Energy Harvesting. Proc. of the 2011-14th European Conference on Power Electronics and Application (EPE 2011) : 1–10.
67
78. Vo, M.N & Noras, M.A (2013) Energy Harvesting from Electromagnetic Field Surrounding A Current Carrying Conductor. Proc. ESA Annual Meeting on Electrostatics 2013: 1–12.
79. Brogioli D (2009) Extracting renewable energy from a salinity difference using a capacitor. Phys. Rev. Lett. 103 058501–1–4.
80. Guo, W Cao, L Xia, J Nie, F-Q Ma, W Xue, J Song,Y Zhu, D Wang, Y & Jian L (2010) Energy Harvesting with Single-Ion Selective Nanopores: A Concentration –Gradient-Driven Nanofluidic Power Source. Advanced Functional Materials 20(8):1339–44.
81. Leonov V, Fiorini P, Sedky S, Torfs T & Van Hoof C (2005) Thermoelectric MEMS generators as a power supply for a body area network. In: Proceedings of the 13th int conf on solid–state sensors, actuators and microsystems, transducers 2005: 291–294.
82. Kishi, M Nemoto, H Hamao, T Yamamoto, M Sudou, S Mandai, M & Yamamoto, S. (1999) Micro thermoelectric modules and their application to wristwatches as an energy source. In: 18th International conference on thermoelectrics: 301–7.
83. Strasser, M Aigner, R Lauterbach, C Sturm, T.F. Franosch, M & Wachutka, G (2004) Micromachined CMOS thermoelectric generator as on-chip power supply. Sensors and Actuators A 114: 362–70.
84. Ferrari M, Ferrari V, Guizzetti M, Marioli D & Taroni A (2008) Piezoelectric Multifrequency Energy Converter For Power Harvesting In Autonomous Microsystems Sensors and Actuators A 142(1): 329–335.
85. Kim, I.H Jung, H.J Lee, B.M & Jang, S.J (2011) Broadband energy-harvesting using a two degree-of-freedom vibrating body. Applied Phys. Lett. 98:1–3.
86. Lin, S.C Lee, B.S Wu, W.J & Lee, C.K (2009) Multi-cantilever piezoelectric MEMS generator in energy harvesting. IEEE Int. Ultrason. Symp. Proc., 1: 755–758.
87. G.H. Feng, J.C. Hung (2007) Optimal FOM designed piezoelectric microgenerator with energy harvesting in a wide vibration bandwidth. Proc. 2nd IEEE Int. Conf. on Nano/Micro Engineered and Molecular Systems (16–19) January 2007, Bangkok, Thailand) pp. 511–514.
88. Roundy S, Leland S E, Baker J, Carleton E, Reilly E, Lai E, Otis B, Rabaey J M, Sundararajan V & Wright PK (2005) Improving power output for vibration-based energy scavengers, Pervasive. Comput. 4(1): 28–36.
89. Ferrari M, Ferrari V, Guizzetti M, Andò B, Baglio S & Trigona C (2010) Improved energy harvesting from wideband vibrations by nonlinear piezoelectric converters Sensors and Actuators A: Physical 162(2): 425–31.
90. Cottone F, Gammaitoni L, Vocca H, Ferrari M & Ferrari V (2012) Piezoelectric buckled beamsfor randomvibration energy harvesting. Smart Materials and Structures 21 035021.
91. Liu WQ, Badel A, Formosa F, Wu Y & Agbossou A (2013) Novel piezoelectric bistable oscillator architecture for wideband vibration energy harvesting. Smart Materials and Structures 22(3) 035013.
68
92. Xue, H Hu, Y & Wang, Q.M (2008) Broadband piezoelectric energy harvesting devices using multiple bimorphs with different operating frequencies. IEEE T. Ultrason. Ferr., 55: 2104–8
93. Shahruz S.M (2006) Design of mechanical band-pass filters for energy scavenging. J. Sound Vib. 292: 987–998.
94. Morgan Electro Ceramics Inc. (2013) material datasheet URI: http://www.morganelectroceramics.com Cited 2013/10/12
95. Mitcheson, P. D Reilly, E. K Toh, T Wright, P. K & Yeatman, E. M. Performance limits of the three MEMS inertial energy generator transduction types. J. Micromech. Microeng., 2007, 17(9): 211–6.
96. Howells CA (2008) Piezoelectric Energy Conversion using Locomotion. American Society of Mechanical Engineers, 2nd International Conference on Energy Sustainability ES2008-54063: 13–17.
97. Rocha JG, Goncalves LM, Rocha PF, Silva MP & Lanceros-Mendez S (2010) Energy harvesting from piezoelectric materials fully integrated in footwear. IEEE Transactions on Industrial Electronics 57(3): 813–19.
98. Pozzi M, Aung MSH, Zhu M, Jones RK & Goulermas JY (2012) The pizzicato knee-joint energy harvester: characterization with biomechanical data and the effect of backpack load. Smart Materials and Structures 21(7)
99. Feenstra J, Granstrom J & Sodano H (2008) Energy harvesting through backpack employing a mechanically amplified piezoelectric stack. Mechanical systems and Signal processing 22(3): 721-34.
100. S.E. Jo, M.K. Kim, M.S. Kim and Y.J. Kim (2012) Flexible thermoelectric generator for human body heat energy harvesting, Electronics letters, 48(16)
101. Dogan, A Fernandez, J.F Uchino, K & Newnham, R.E (1996) The “cymbal” ™ electromechanical actuator, Proc. 10th IEEE Symp. Appl. Ferroelect. 213–6.
102. Lam, K.H Wang, X.X & Chan, H.L.W (2005) Lead-free piezoceramic cymbal actuator. Sensors and Actuators A: Physical 125(2): 393–7.
103. Meyer, R.J Hughes, W.J Montgomery, T,C Markley, D.C & Newnham R.E (2002) Design of fabrication improvements to the cymbal transducer aided by finite element analysis. Journal of Electroceramics 8: 163–74.
104. Ochoa, P Villegas, M Pons, J.L Leidinger, P & Fernández, J.F (2005) Tunability of cymbals as piezocomposite transducers. Journal of Electroceramics 14: 221–9.
105. Tressler, J.F Cao, W Uchino, K & Newnham R.E (1998) Finite element analysis of the cymbal-type flextensional transducer. IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control 45(5): 1363–9.
106. Narayanan, M & Schwartz RW (2010) Design, fabrication and finite element modelling of a new wagon wheel flextensional transducer. Journal of Electroceramics 24(3): 205–13.
107. Wang, B & Xia, Q (2008) Numerical analysis on two kinds of cymbal membranes for piezoelectric micro-flow actuator. Journal of Intelligent Material Systems and Structures 19(3): 343–9
69
108. Kim, H.W Priya, S Uchino, K & R,E Newnham (2005) Piezoelectric energy harvesting under high pre-stressed cyclic vibrations. Journal of Electroceramics 15: 27–34.
109. Kim, H. W Batra, A Priya, S Uchino, K., Markley, D., Newnham, R. E & Hofmann, H (2004) Energy Harvesting Using a Piezoelectric “Cymbal” Transducer in Dynamic Environment. Japanese Journal of Applied Physics 43: 6178–83.
110. Sohn, J.W Choi, S.B & Lee, D.Y (2005) An investigation on piezoelectric energy harvesting for MEMS power sources. Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science 219(4): 429–436.
111. Mo, C Arnold, D Kinsel, W.C & Clark W.W (2013) Modeling and experimental validation of unimorph piezoelectric cymbal design in energy harvesting. Journal of Intelligent Material Systems and Structures. 24(7): 828–36.
112. Yuan, J Shan, X Xie, T & Chen, W (2010) Modeling and improvement of a cymbal transducer in energy harvesting. Journal of Intelligent Material Systems and Structures 21(8): 765–71.
113. Živanović S, Racić V, El-Bahnasy I & Pavić A (2007) Statistical characterisation of parameters defining human walking as observed on an indoor passerelle. Experimental vibration analysis for civil engineering structures: 219–225.
114. Matsumoto Y, Nishioka T, Shiojiri H & Matsuzaki K (1978) Dynamic design of footbridges. IABSE proceedings P(17): 1–15.
115. Pachi, A & Ji, T (2005) Frequency and velocity of people walking. Struct Eng, 83(3): 36–40.
71
Original papers
I Leinonen M, Juuti J & Jantunen H (2005) Equivalent circuit based modification of a pre-stressed piezo actuator. Proc. 4th Int. Conf. on Smart Systems, Seinäjoki.
II Sobocinski M, Leinonen M, Juuti J & Jantunen H (2012) A Piezoelectric Active Mirror Suspension System Embedded Into Low-Temperature Cofired Ceramic. IEEE trans. on Ultrasonics, Ferroelectrics and frequency control 59(9): 1990–1995.
III Leinonen M, Palosaari J, Juuti J & Jantunen H (2011) Piezoelectric energy harvester for vibrating environments using multiple beam topology for wideband operation. AP XIX Proceedings 41, Helsinki.
IV Sobocinski M, Leinonen M, Juuti J & Jantunen H (2011) Monomorph piezoelectric wideband energy harvester integrated into LTCC. Journal Of European Ceramic Society 31(5): 789–794.
V Leinonen M, Palosaari J, Juuti J & Jantunen H (2014) Combined electrical and electromechanical simulations of a piezoelectric Cymbal harvester for energy harvesting from walking. Journal of Intelligent Material Systems and Structures, 25(4): 391–400.
Reprinted with permission from Frami OY (Paper I), IEEE (Paper II), Suomen
automaatioseura (Paper III), Elsevier Ltd (Paper IV) and Sage publishing (V).
Original publications are not included in the electronic version of the dissertation.
A C T A U N I V E R S I T A T I S O U L U E N S I S
Book orders:Granum: Virtual book storehttp://granum.uta.fi/granum/
S E R I E S C T E C H N I C A
514. Sobocinski, Maciej (2014) Embedding of bulk piezoelectric structures in LowTemperature Co-fired Ceramic
515. Kulju, Timo (2014) Utilization of phenomena-based modeling in unit operationdesign
516. Karinkanta, Pasi (2014) Dry fine grinding of Norway spruce (Picea abies) wood inimpact-based fine grinding mills
517. Tervo, Valtteri (2015) Joint multiuser power allocation and iterative multi-antenna receiver design
518. Jayasinghe, Laddu Keeth Saliya (2015) Analysis on MIMO relaying scenarios inwireless communication systems
519. Partala, Juha (2015) Algebraic methods for cryptographic key exhange
520. Karvonen, Heikki (2015) Energy efficiency improvements for wireless sensornetworks by using cross-layer analysis
521. Putaala, Jussi (2015) Reliability and prognostic monitoring methods of electronicsinterconnections in advanced SMD applications
522. Pirilä, Minna (2015) Adsorption and photocatalysis in water treatment : active,abundant and inexpensive materials and methods
523. Alves, Hirley (2015) On the performance analysis of full-duplex networks
524. Siirtola, Pekka (2015) Recognizing human activities based on wearable inertialmeasurements : methods and applications
525. Lu, Pen-Shun (2015) Decoding and lossy forwarding based multiple accessrelaying
526. Suopajärvi, Terhi (2015) Functionalized nanocelluloses in wastewater treatmentapplications
527. Pekuri, Aki (2015) The role of business models in construction businessmanagement
528. Mantere, Matti (2015) Network security monitoring and anomaly detection inindustrial control system networks
529. Piri, Esa (2015) Improving heterogeneous wireless networking with cross-layerinformation services
530. Leppänen, Kimmo (2015) Sample preparation method and synchronizedthermography to characterize uniformity of conductive thin films
C532etukansi.kesken.fm Page 2 Tuesday, May 12, 2015 12:55 PM
UNIVERSITY OF OULU P .O. Box 8000 F I -90014 UNIVERSITY OF OULU FINLAND
A C T A U N I V E R S I T A T I S O U L U E N S I S
Professor Esa Hohtola
University Lecturer Santeri Palviainen
Postdoctoral research fellow Sanna Taskila
Professor Olli Vuolteenaho
University Lecturer Veli-Matti Ulvinen
Director Sinikka Eskelinen
Professor Jari Juga
University Lecturer Anu Soikkeli
Professor Olli Vuolteenaho
Publications Editor Kirsti Nurkkala
ISBN 978-952-62-0836-7 (Paperback)ISBN 978-952-62-0837-4 (PDF)ISSN 0355-3213 (Print)ISSN 1796-2226 (Online)
U N I V E R S I TAT I S O U L U E N S I SACTAC
TECHNICA
U N I V E R S I TAT I S O U L U E N S I SACTAC
TECHNICA
OULU 2015
C 532
Mikko Leinonen
FINITE ELEMENT METHOD AND EQUIVALENT CIRCUIT BASED DESIGN OF PIEZOELECTRIC ACTUATORS AND ENERGY HARVESTER DYNAMICS
UNIVERSITY OF OULU GRADUATE SCHOOL;UNIVERSITY OF OULU,FACULTY OF INFORMATION TECHNOLOGY AND ELECTRICAL ENGINEERING,DEPARTMENT OF ELECTRICAL ENGINEERING
C 532
ACTA
Mikko Leinonen
C532etukansi.kesken.fm Page 1 Tuesday, May 12, 2015 12:55 PM